at v6.5 270 kB view raw
1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8#ifndef _UAPI__LINUX_BPF_H__ 9#define _UAPI__LINUX_BPF_H__ 10 11#include <linux/types.h> 12#include <linux/bpf_common.h> 13 14/* Extended instruction set based on top of classic BPF */ 15 16/* instruction classes */ 17#define BPF_JMP32 0x06 /* jmp mode in word width */ 18#define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20/* ld/ldx fields */ 21#define BPF_DW 0x18 /* double word (64-bit) */ 22#define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23#define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25/* alu/jmp fields */ 26#define BPF_MOV 0xb0 /* mov reg to reg */ 27#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29/* change endianness of a register */ 30#define BPF_END 0xd0 /* flags for endianness conversion: */ 31#define BPF_TO_LE 0x00 /* convert to little-endian */ 32#define BPF_TO_BE 0x08 /* convert to big-endian */ 33#define BPF_FROM_LE BPF_TO_LE 34#define BPF_FROM_BE BPF_TO_BE 35 36/* jmp encodings */ 37#define BPF_JNE 0x50 /* jump != */ 38#define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42#define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44#define BPF_CALL 0x80 /* function call */ 45#define BPF_EXIT 0x90 /* function return */ 46 47/* atomic op type fields (stored in immediate) */ 48#define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49#define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50#define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52/* Register numbers */ 53enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66}; 67 68/* BPF has 10 general purpose 64-bit registers and stack frame. */ 69#define MAX_BPF_REG __MAX_BPF_REG 70 71struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77}; 78 79/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83}; 84 85struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88}; 89 90enum bpf_cgroup_iter_order { 91 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 92 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 93 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 94 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 95 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 96}; 97 98union bpf_iter_link_info { 99 struct { 100 __u32 map_fd; 101 } map; 102 struct { 103 enum bpf_cgroup_iter_order order; 104 105 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 106 * both are zero, the walk starts from the default cgroup v2 107 * root. For walking v1 hierarchy, one should always explicitly 108 * specify cgroup_fd. 109 */ 110 __u32 cgroup_fd; 111 __u64 cgroup_id; 112 } cgroup; 113 /* Parameters of task iterators. */ 114 struct { 115 __u32 tid; 116 __u32 pid; 117 __u32 pid_fd; 118 } task; 119}; 120 121/* BPF syscall commands, see bpf(2) man-page for more details. */ 122/** 123 * DOC: eBPF Syscall Preamble 124 * 125 * The operation to be performed by the **bpf**\ () system call is determined 126 * by the *cmd* argument. Each operation takes an accompanying argument, 127 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 128 * below). The size argument is the size of the union pointed to by *attr*. 129 */ 130/** 131 * DOC: eBPF Syscall Commands 132 * 133 * BPF_MAP_CREATE 134 * Description 135 * Create a map and return a file descriptor that refers to the 136 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 137 * is automatically enabled for the new file descriptor. 138 * 139 * Applying **close**\ (2) to the file descriptor returned by 140 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 141 * 142 * Return 143 * A new file descriptor (a nonnegative integer), or -1 if an 144 * error occurred (in which case, *errno* is set appropriately). 145 * 146 * BPF_MAP_LOOKUP_ELEM 147 * Description 148 * Look up an element with a given *key* in the map referred to 149 * by the file descriptor *map_fd*. 150 * 151 * The *flags* argument may be specified as one of the 152 * following: 153 * 154 * **BPF_F_LOCK** 155 * Look up the value of a spin-locked map without 156 * returning the lock. This must be specified if the 157 * elements contain a spinlock. 158 * 159 * Return 160 * Returns zero on success. On error, -1 is returned and *errno* 161 * is set appropriately. 162 * 163 * BPF_MAP_UPDATE_ELEM 164 * Description 165 * Create or update an element (key/value pair) in a specified map. 166 * 167 * The *flags* argument should be specified as one of the 168 * following: 169 * 170 * **BPF_ANY** 171 * Create a new element or update an existing element. 172 * **BPF_NOEXIST** 173 * Create a new element only if it did not exist. 174 * **BPF_EXIST** 175 * Update an existing element. 176 * **BPF_F_LOCK** 177 * Update a spin_lock-ed map element. 178 * 179 * Return 180 * Returns zero on success. On error, -1 is returned and *errno* 181 * is set appropriately. 182 * 183 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 184 * **E2BIG**, **EEXIST**, or **ENOENT**. 185 * 186 * **E2BIG** 187 * The number of elements in the map reached the 188 * *max_entries* limit specified at map creation time. 189 * **EEXIST** 190 * If *flags* specifies **BPF_NOEXIST** and the element 191 * with *key* already exists in the map. 192 * **ENOENT** 193 * If *flags* specifies **BPF_EXIST** and the element with 194 * *key* does not exist in the map. 195 * 196 * BPF_MAP_DELETE_ELEM 197 * Description 198 * Look up and delete an element by key in a specified map. 199 * 200 * Return 201 * Returns zero on success. On error, -1 is returned and *errno* 202 * is set appropriately. 203 * 204 * BPF_MAP_GET_NEXT_KEY 205 * Description 206 * Look up an element by key in a specified map and return the key 207 * of the next element. Can be used to iterate over all elements 208 * in the map. 209 * 210 * Return 211 * Returns zero on success. On error, -1 is returned and *errno* 212 * is set appropriately. 213 * 214 * The following cases can be used to iterate over all elements of 215 * the map: 216 * 217 * * If *key* is not found, the operation returns zero and sets 218 * the *next_key* pointer to the key of the first element. 219 * * If *key* is found, the operation returns zero and sets the 220 * *next_key* pointer to the key of the next element. 221 * * If *key* is the last element, returns -1 and *errno* is set 222 * to **ENOENT**. 223 * 224 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 225 * **EINVAL** on error. 226 * 227 * BPF_PROG_LOAD 228 * Description 229 * Verify and load an eBPF program, returning a new file 230 * descriptor associated with the program. 231 * 232 * Applying **close**\ (2) to the file descriptor returned by 233 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 234 * 235 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 236 * automatically enabled for the new file descriptor. 237 * 238 * Return 239 * A new file descriptor (a nonnegative integer), or -1 if an 240 * error occurred (in which case, *errno* is set appropriately). 241 * 242 * BPF_OBJ_PIN 243 * Description 244 * Pin an eBPF program or map referred by the specified *bpf_fd* 245 * to the provided *pathname* on the filesystem. 246 * 247 * The *pathname* argument must not contain a dot ("."). 248 * 249 * On success, *pathname* retains a reference to the eBPF object, 250 * preventing deallocation of the object when the original 251 * *bpf_fd* is closed. This allow the eBPF object to live beyond 252 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 253 * process. 254 * 255 * Applying **unlink**\ (2) or similar calls to the *pathname* 256 * unpins the object from the filesystem, removing the reference. 257 * If no other file descriptors or filesystem nodes refer to the 258 * same object, it will be deallocated (see NOTES). 259 * 260 * The filesystem type for the parent directory of *pathname* must 261 * be **BPF_FS_MAGIC**. 262 * 263 * Return 264 * Returns zero on success. On error, -1 is returned and *errno* 265 * is set appropriately. 266 * 267 * BPF_OBJ_GET 268 * Description 269 * Open a file descriptor for the eBPF object pinned to the 270 * specified *pathname*. 271 * 272 * Return 273 * A new file descriptor (a nonnegative integer), or -1 if an 274 * error occurred (in which case, *errno* is set appropriately). 275 * 276 * BPF_PROG_ATTACH 277 * Description 278 * Attach an eBPF program to a *target_fd* at the specified 279 * *attach_type* hook. 280 * 281 * The *attach_type* specifies the eBPF attachment point to 282 * attach the program to, and must be one of *bpf_attach_type* 283 * (see below). 284 * 285 * The *attach_bpf_fd* must be a valid file descriptor for a 286 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 287 * or sock_ops type corresponding to the specified *attach_type*. 288 * 289 * The *target_fd* must be a valid file descriptor for a kernel 290 * object which depends on the attach type of *attach_bpf_fd*: 291 * 292 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 293 * **BPF_PROG_TYPE_CGROUP_SKB**, 294 * **BPF_PROG_TYPE_CGROUP_SOCK**, 295 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 296 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 297 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 298 * **BPF_PROG_TYPE_SOCK_OPS** 299 * 300 * Control Group v2 hierarchy with the eBPF controller 301 * enabled. Requires the kernel to be compiled with 302 * **CONFIG_CGROUP_BPF**. 303 * 304 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 305 * 306 * Network namespace (eg /proc/self/ns/net). 307 * 308 * **BPF_PROG_TYPE_LIRC_MODE2** 309 * 310 * LIRC device path (eg /dev/lircN). Requires the kernel 311 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 312 * 313 * **BPF_PROG_TYPE_SK_SKB**, 314 * **BPF_PROG_TYPE_SK_MSG** 315 * 316 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 317 * 318 * Return 319 * Returns zero on success. On error, -1 is returned and *errno* 320 * is set appropriately. 321 * 322 * BPF_PROG_DETACH 323 * Description 324 * Detach the eBPF program associated with the *target_fd* at the 325 * hook specified by *attach_type*. The program must have been 326 * previously attached using **BPF_PROG_ATTACH**. 327 * 328 * Return 329 * Returns zero on success. On error, -1 is returned and *errno* 330 * is set appropriately. 331 * 332 * BPF_PROG_TEST_RUN 333 * Description 334 * Run the eBPF program associated with the *prog_fd* a *repeat* 335 * number of times against a provided program context *ctx_in* and 336 * data *data_in*, and return the modified program context 337 * *ctx_out*, *data_out* (for example, packet data), result of the 338 * execution *retval*, and *duration* of the test run. 339 * 340 * The sizes of the buffers provided as input and output 341 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 342 * be provided in the corresponding variables *ctx_size_in*, 343 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 344 * of these parameters are not provided (ie set to NULL), the 345 * corresponding size field must be zero. 346 * 347 * Some program types have particular requirements: 348 * 349 * **BPF_PROG_TYPE_SK_LOOKUP** 350 * *data_in* and *data_out* must be NULL. 351 * 352 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 353 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 354 * 355 * *ctx_out*, *data_in* and *data_out* must be NULL. 356 * *repeat* must be zero. 357 * 358 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 359 * 360 * Return 361 * Returns zero on success. On error, -1 is returned and *errno* 362 * is set appropriately. 363 * 364 * **ENOSPC** 365 * Either *data_size_out* or *ctx_size_out* is too small. 366 * **ENOTSUPP** 367 * This command is not supported by the program type of 368 * the program referred to by *prog_fd*. 369 * 370 * BPF_PROG_GET_NEXT_ID 371 * Description 372 * Fetch the next eBPF program currently loaded into the kernel. 373 * 374 * Looks for the eBPF program with an id greater than *start_id* 375 * and updates *next_id* on success. If no other eBPF programs 376 * remain with ids higher than *start_id*, returns -1 and sets 377 * *errno* to **ENOENT**. 378 * 379 * Return 380 * Returns zero on success. On error, or when no id remains, -1 381 * is returned and *errno* is set appropriately. 382 * 383 * BPF_MAP_GET_NEXT_ID 384 * Description 385 * Fetch the next eBPF map currently loaded into the kernel. 386 * 387 * Looks for the eBPF map with an id greater than *start_id* 388 * and updates *next_id* on success. If no other eBPF maps 389 * remain with ids higher than *start_id*, returns -1 and sets 390 * *errno* to **ENOENT**. 391 * 392 * Return 393 * Returns zero on success. On error, or when no id remains, -1 394 * is returned and *errno* is set appropriately. 395 * 396 * BPF_PROG_GET_FD_BY_ID 397 * Description 398 * Open a file descriptor for the eBPF program corresponding to 399 * *prog_id*. 400 * 401 * Return 402 * A new file descriptor (a nonnegative integer), or -1 if an 403 * error occurred (in which case, *errno* is set appropriately). 404 * 405 * BPF_MAP_GET_FD_BY_ID 406 * Description 407 * Open a file descriptor for the eBPF map corresponding to 408 * *map_id*. 409 * 410 * Return 411 * A new file descriptor (a nonnegative integer), or -1 if an 412 * error occurred (in which case, *errno* is set appropriately). 413 * 414 * BPF_OBJ_GET_INFO_BY_FD 415 * Description 416 * Obtain information about the eBPF object corresponding to 417 * *bpf_fd*. 418 * 419 * Populates up to *info_len* bytes of *info*, which will be in 420 * one of the following formats depending on the eBPF object type 421 * of *bpf_fd*: 422 * 423 * * **struct bpf_prog_info** 424 * * **struct bpf_map_info** 425 * * **struct bpf_btf_info** 426 * * **struct bpf_link_info** 427 * 428 * Return 429 * Returns zero on success. On error, -1 is returned and *errno* 430 * is set appropriately. 431 * 432 * BPF_PROG_QUERY 433 * Description 434 * Obtain information about eBPF programs associated with the 435 * specified *attach_type* hook. 436 * 437 * The *target_fd* must be a valid file descriptor for a kernel 438 * object which depends on the attach type of *attach_bpf_fd*: 439 * 440 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 441 * **BPF_PROG_TYPE_CGROUP_SKB**, 442 * **BPF_PROG_TYPE_CGROUP_SOCK**, 443 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 444 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 445 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 446 * **BPF_PROG_TYPE_SOCK_OPS** 447 * 448 * Control Group v2 hierarchy with the eBPF controller 449 * enabled. Requires the kernel to be compiled with 450 * **CONFIG_CGROUP_BPF**. 451 * 452 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 453 * 454 * Network namespace (eg /proc/self/ns/net). 455 * 456 * **BPF_PROG_TYPE_LIRC_MODE2** 457 * 458 * LIRC device path (eg /dev/lircN). Requires the kernel 459 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 460 * 461 * **BPF_PROG_QUERY** always fetches the number of programs 462 * attached and the *attach_flags* which were used to attach those 463 * programs. Additionally, if *prog_ids* is nonzero and the number 464 * of attached programs is less than *prog_cnt*, populates 465 * *prog_ids* with the eBPF program ids of the programs attached 466 * at *target_fd*. 467 * 468 * The following flags may alter the result: 469 * 470 * **BPF_F_QUERY_EFFECTIVE** 471 * Only return information regarding programs which are 472 * currently effective at the specified *target_fd*. 473 * 474 * Return 475 * Returns zero on success. On error, -1 is returned and *errno* 476 * is set appropriately. 477 * 478 * BPF_RAW_TRACEPOINT_OPEN 479 * Description 480 * Attach an eBPF program to a tracepoint *name* to access kernel 481 * internal arguments of the tracepoint in their raw form. 482 * 483 * The *prog_fd* must be a valid file descriptor associated with 484 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 485 * 486 * No ABI guarantees are made about the content of tracepoint 487 * arguments exposed to the corresponding eBPF program. 488 * 489 * Applying **close**\ (2) to the file descriptor returned by 490 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 491 * 492 * Return 493 * A new file descriptor (a nonnegative integer), or -1 if an 494 * error occurred (in which case, *errno* is set appropriately). 495 * 496 * BPF_BTF_LOAD 497 * Description 498 * Verify and load BPF Type Format (BTF) metadata into the kernel, 499 * returning a new file descriptor associated with the metadata. 500 * BTF is described in more detail at 501 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 502 * 503 * The *btf* parameter must point to valid memory providing 504 * *btf_size* bytes of BTF binary metadata. 505 * 506 * The returned file descriptor can be passed to other **bpf**\ () 507 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 508 * associate the BTF with those objects. 509 * 510 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 511 * parameters to specify a *btf_log_buf*, *btf_log_size* and 512 * *btf_log_level* which allow the kernel to return freeform log 513 * output regarding the BTF verification process. 514 * 515 * Return 516 * A new file descriptor (a nonnegative integer), or -1 if an 517 * error occurred (in which case, *errno* is set appropriately). 518 * 519 * BPF_BTF_GET_FD_BY_ID 520 * Description 521 * Open a file descriptor for the BPF Type Format (BTF) 522 * corresponding to *btf_id*. 523 * 524 * Return 525 * A new file descriptor (a nonnegative integer), or -1 if an 526 * error occurred (in which case, *errno* is set appropriately). 527 * 528 * BPF_TASK_FD_QUERY 529 * Description 530 * Obtain information about eBPF programs associated with the 531 * target process identified by *pid* and *fd*. 532 * 533 * If the *pid* and *fd* are associated with a tracepoint, kprobe 534 * or uprobe perf event, then the *prog_id* and *fd_type* will 535 * be populated with the eBPF program id and file descriptor type 536 * of type **bpf_task_fd_type**. If associated with a kprobe or 537 * uprobe, the *probe_offset* and *probe_addr* will also be 538 * populated. Optionally, if *buf* is provided, then up to 539 * *buf_len* bytes of *buf* will be populated with the name of 540 * the tracepoint, kprobe or uprobe. 541 * 542 * The resulting *prog_id* may be introspected in deeper detail 543 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 544 * 545 * Return 546 * Returns zero on success. On error, -1 is returned and *errno* 547 * is set appropriately. 548 * 549 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 550 * Description 551 * Look up an element with the given *key* in the map referred to 552 * by the file descriptor *fd*, and if found, delete the element. 553 * 554 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 555 * types, the *flags* argument needs to be set to 0, but for other 556 * map types, it may be specified as: 557 * 558 * **BPF_F_LOCK** 559 * Look up and delete the value of a spin-locked map 560 * without returning the lock. This must be specified if 561 * the elements contain a spinlock. 562 * 563 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 564 * implement this command as a "pop" operation, deleting the top 565 * element rather than one corresponding to *key*. 566 * The *key* and *key_len* parameters should be zeroed when 567 * issuing this operation for these map types. 568 * 569 * This command is only valid for the following map types: 570 * * **BPF_MAP_TYPE_QUEUE** 571 * * **BPF_MAP_TYPE_STACK** 572 * * **BPF_MAP_TYPE_HASH** 573 * * **BPF_MAP_TYPE_PERCPU_HASH** 574 * * **BPF_MAP_TYPE_LRU_HASH** 575 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 576 * 577 * Return 578 * Returns zero on success. On error, -1 is returned and *errno* 579 * is set appropriately. 580 * 581 * BPF_MAP_FREEZE 582 * Description 583 * Freeze the permissions of the specified map. 584 * 585 * Write permissions may be frozen by passing zero *flags*. 586 * Upon success, no future syscall invocations may alter the 587 * map state of *map_fd*. Write operations from eBPF programs 588 * are still possible for a frozen map. 589 * 590 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 591 * 592 * Return 593 * Returns zero on success. On error, -1 is returned and *errno* 594 * is set appropriately. 595 * 596 * BPF_BTF_GET_NEXT_ID 597 * Description 598 * Fetch the next BPF Type Format (BTF) object currently loaded 599 * into the kernel. 600 * 601 * Looks for the BTF object with an id greater than *start_id* 602 * and updates *next_id* on success. If no other BTF objects 603 * remain with ids higher than *start_id*, returns -1 and sets 604 * *errno* to **ENOENT**. 605 * 606 * Return 607 * Returns zero on success. On error, or when no id remains, -1 608 * is returned and *errno* is set appropriately. 609 * 610 * BPF_MAP_LOOKUP_BATCH 611 * Description 612 * Iterate and fetch multiple elements in a map. 613 * 614 * Two opaque values are used to manage batch operations, 615 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 616 * to NULL to begin the batched operation. After each subsequent 617 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 618 * *out_batch* as the *in_batch* for the next operation to 619 * continue iteration from the current point. 620 * 621 * The *keys* and *values* are output parameters which must point 622 * to memory large enough to hold *count* items based on the key 623 * and value size of the map *map_fd*. The *keys* buffer must be 624 * of *key_size* * *count*. The *values* buffer must be of 625 * *value_size* * *count*. 626 * 627 * The *elem_flags* argument may be specified as one of the 628 * following: 629 * 630 * **BPF_F_LOCK** 631 * Look up the value of a spin-locked map without 632 * returning the lock. This must be specified if the 633 * elements contain a spinlock. 634 * 635 * On success, *count* elements from the map are copied into the 636 * user buffer, with the keys copied into *keys* and the values 637 * copied into the corresponding indices in *values*. 638 * 639 * If an error is returned and *errno* is not **EFAULT**, *count* 640 * is set to the number of successfully processed elements. 641 * 642 * Return 643 * Returns zero on success. On error, -1 is returned and *errno* 644 * is set appropriately. 645 * 646 * May set *errno* to **ENOSPC** to indicate that *keys* or 647 * *values* is too small to dump an entire bucket during 648 * iteration of a hash-based map type. 649 * 650 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 651 * Description 652 * Iterate and delete all elements in a map. 653 * 654 * This operation has the same behavior as 655 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 656 * 657 * * Every element that is successfully returned is also deleted 658 * from the map. This is at least *count* elements. Note that 659 * *count* is both an input and an output parameter. 660 * * Upon returning with *errno* set to **EFAULT**, up to 661 * *count* elements may be deleted without returning the keys 662 * and values of the deleted elements. 663 * 664 * Return 665 * Returns zero on success. On error, -1 is returned and *errno* 666 * is set appropriately. 667 * 668 * BPF_MAP_UPDATE_BATCH 669 * Description 670 * Update multiple elements in a map by *key*. 671 * 672 * The *keys* and *values* are input parameters which must point 673 * to memory large enough to hold *count* items based on the key 674 * and value size of the map *map_fd*. The *keys* buffer must be 675 * of *key_size* * *count*. The *values* buffer must be of 676 * *value_size* * *count*. 677 * 678 * Each element specified in *keys* is sequentially updated to the 679 * value in the corresponding index in *values*. The *in_batch* 680 * and *out_batch* parameters are ignored and should be zeroed. 681 * 682 * The *elem_flags* argument should be specified as one of the 683 * following: 684 * 685 * **BPF_ANY** 686 * Create new elements or update a existing elements. 687 * **BPF_NOEXIST** 688 * Create new elements only if they do not exist. 689 * **BPF_EXIST** 690 * Update existing elements. 691 * **BPF_F_LOCK** 692 * Update spin_lock-ed map elements. This must be 693 * specified if the map value contains a spinlock. 694 * 695 * On success, *count* elements from the map are updated. 696 * 697 * If an error is returned and *errno* is not **EFAULT**, *count* 698 * is set to the number of successfully processed elements. 699 * 700 * Return 701 * Returns zero on success. On error, -1 is returned and *errno* 702 * is set appropriately. 703 * 704 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 705 * **E2BIG**. **E2BIG** indicates that the number of elements in 706 * the map reached the *max_entries* limit specified at map 707 * creation time. 708 * 709 * May set *errno* to one of the following error codes under 710 * specific circumstances: 711 * 712 * **EEXIST** 713 * If *flags* specifies **BPF_NOEXIST** and the element 714 * with *key* already exists in the map. 715 * **ENOENT** 716 * If *flags* specifies **BPF_EXIST** and the element with 717 * *key* does not exist in the map. 718 * 719 * BPF_MAP_DELETE_BATCH 720 * Description 721 * Delete multiple elements in a map by *key*. 722 * 723 * The *keys* parameter is an input parameter which must point 724 * to memory large enough to hold *count* items based on the key 725 * size of the map *map_fd*, that is, *key_size* * *count*. 726 * 727 * Each element specified in *keys* is sequentially deleted. The 728 * *in_batch*, *out_batch*, and *values* parameters are ignored 729 * and should be zeroed. 730 * 731 * The *elem_flags* argument may be specified as one of the 732 * following: 733 * 734 * **BPF_F_LOCK** 735 * Look up the value of a spin-locked map without 736 * returning the lock. This must be specified if the 737 * elements contain a spinlock. 738 * 739 * On success, *count* elements from the map are updated. 740 * 741 * If an error is returned and *errno* is not **EFAULT**, *count* 742 * is set to the number of successfully processed elements. If 743 * *errno* is **EFAULT**, up to *count* elements may be been 744 * deleted. 745 * 746 * Return 747 * Returns zero on success. On error, -1 is returned and *errno* 748 * is set appropriately. 749 * 750 * BPF_LINK_CREATE 751 * Description 752 * Attach an eBPF program to a *target_fd* at the specified 753 * *attach_type* hook and return a file descriptor handle for 754 * managing the link. 755 * 756 * Return 757 * A new file descriptor (a nonnegative integer), or -1 if an 758 * error occurred (in which case, *errno* is set appropriately). 759 * 760 * BPF_LINK_UPDATE 761 * Description 762 * Update the eBPF program in the specified *link_fd* to 763 * *new_prog_fd*. 764 * 765 * Return 766 * Returns zero on success. On error, -1 is returned and *errno* 767 * is set appropriately. 768 * 769 * BPF_LINK_GET_FD_BY_ID 770 * Description 771 * Open a file descriptor for the eBPF Link corresponding to 772 * *link_id*. 773 * 774 * Return 775 * A new file descriptor (a nonnegative integer), or -1 if an 776 * error occurred (in which case, *errno* is set appropriately). 777 * 778 * BPF_LINK_GET_NEXT_ID 779 * Description 780 * Fetch the next eBPF link currently loaded into the kernel. 781 * 782 * Looks for the eBPF link with an id greater than *start_id* 783 * and updates *next_id* on success. If no other eBPF links 784 * remain with ids higher than *start_id*, returns -1 and sets 785 * *errno* to **ENOENT**. 786 * 787 * Return 788 * Returns zero on success. On error, or when no id remains, -1 789 * is returned and *errno* is set appropriately. 790 * 791 * BPF_ENABLE_STATS 792 * Description 793 * Enable eBPF runtime statistics gathering. 794 * 795 * Runtime statistics gathering for the eBPF runtime is disabled 796 * by default to minimize the corresponding performance overhead. 797 * This command enables statistics globally. 798 * 799 * Multiple programs may independently enable statistics. 800 * After gathering the desired statistics, eBPF runtime statistics 801 * may be disabled again by calling **close**\ (2) for the file 802 * descriptor returned by this function. Statistics will only be 803 * disabled system-wide when all outstanding file descriptors 804 * returned by prior calls for this subcommand are closed. 805 * 806 * Return 807 * A new file descriptor (a nonnegative integer), or -1 if an 808 * error occurred (in which case, *errno* is set appropriately). 809 * 810 * BPF_ITER_CREATE 811 * Description 812 * Create an iterator on top of the specified *link_fd* (as 813 * previously created using **BPF_LINK_CREATE**) and return a 814 * file descriptor that can be used to trigger the iteration. 815 * 816 * If the resulting file descriptor is pinned to the filesystem 817 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 818 * for that path will trigger the iterator to read kernel state 819 * using the eBPF program attached to *link_fd*. 820 * 821 * Return 822 * A new file descriptor (a nonnegative integer), or -1 if an 823 * error occurred (in which case, *errno* is set appropriately). 824 * 825 * BPF_LINK_DETACH 826 * Description 827 * Forcefully detach the specified *link_fd* from its 828 * corresponding attachment point. 829 * 830 * Return 831 * Returns zero on success. On error, -1 is returned and *errno* 832 * is set appropriately. 833 * 834 * BPF_PROG_BIND_MAP 835 * Description 836 * Bind a map to the lifetime of an eBPF program. 837 * 838 * The map identified by *map_fd* is bound to the program 839 * identified by *prog_fd* and only released when *prog_fd* is 840 * released. This may be used in cases where metadata should be 841 * associated with a program which otherwise does not contain any 842 * references to the map (for example, embedded in the eBPF 843 * program instructions). 844 * 845 * Return 846 * Returns zero on success. On error, -1 is returned and *errno* 847 * is set appropriately. 848 * 849 * NOTES 850 * eBPF objects (maps and programs) can be shared between processes. 851 * 852 * * After **fork**\ (2), the child inherits file descriptors 853 * referring to the same eBPF objects. 854 * * File descriptors referring to eBPF objects can be transferred over 855 * **unix**\ (7) domain sockets. 856 * * File descriptors referring to eBPF objects can be duplicated in the 857 * usual way, using **dup**\ (2) and similar calls. 858 * * File descriptors referring to eBPF objects can be pinned to the 859 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 860 * 861 * An eBPF object is deallocated only after all file descriptors referring 862 * to the object have been closed and no references remain pinned to the 863 * filesystem or attached (for example, bound to a program or device). 864 */ 865enum bpf_cmd { 866 BPF_MAP_CREATE, 867 BPF_MAP_LOOKUP_ELEM, 868 BPF_MAP_UPDATE_ELEM, 869 BPF_MAP_DELETE_ELEM, 870 BPF_MAP_GET_NEXT_KEY, 871 BPF_PROG_LOAD, 872 BPF_OBJ_PIN, 873 BPF_OBJ_GET, 874 BPF_PROG_ATTACH, 875 BPF_PROG_DETACH, 876 BPF_PROG_TEST_RUN, 877 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 878 BPF_PROG_GET_NEXT_ID, 879 BPF_MAP_GET_NEXT_ID, 880 BPF_PROG_GET_FD_BY_ID, 881 BPF_MAP_GET_FD_BY_ID, 882 BPF_OBJ_GET_INFO_BY_FD, 883 BPF_PROG_QUERY, 884 BPF_RAW_TRACEPOINT_OPEN, 885 BPF_BTF_LOAD, 886 BPF_BTF_GET_FD_BY_ID, 887 BPF_TASK_FD_QUERY, 888 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 889 BPF_MAP_FREEZE, 890 BPF_BTF_GET_NEXT_ID, 891 BPF_MAP_LOOKUP_BATCH, 892 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 893 BPF_MAP_UPDATE_BATCH, 894 BPF_MAP_DELETE_BATCH, 895 BPF_LINK_CREATE, 896 BPF_LINK_UPDATE, 897 BPF_LINK_GET_FD_BY_ID, 898 BPF_LINK_GET_NEXT_ID, 899 BPF_ENABLE_STATS, 900 BPF_ITER_CREATE, 901 BPF_LINK_DETACH, 902 BPF_PROG_BIND_MAP, 903}; 904 905enum bpf_map_type { 906 BPF_MAP_TYPE_UNSPEC, 907 BPF_MAP_TYPE_HASH, 908 BPF_MAP_TYPE_ARRAY, 909 BPF_MAP_TYPE_PROG_ARRAY, 910 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 911 BPF_MAP_TYPE_PERCPU_HASH, 912 BPF_MAP_TYPE_PERCPU_ARRAY, 913 BPF_MAP_TYPE_STACK_TRACE, 914 BPF_MAP_TYPE_CGROUP_ARRAY, 915 BPF_MAP_TYPE_LRU_HASH, 916 BPF_MAP_TYPE_LRU_PERCPU_HASH, 917 BPF_MAP_TYPE_LPM_TRIE, 918 BPF_MAP_TYPE_ARRAY_OF_MAPS, 919 BPF_MAP_TYPE_HASH_OF_MAPS, 920 BPF_MAP_TYPE_DEVMAP, 921 BPF_MAP_TYPE_SOCKMAP, 922 BPF_MAP_TYPE_CPUMAP, 923 BPF_MAP_TYPE_XSKMAP, 924 BPF_MAP_TYPE_SOCKHASH, 925 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 926 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 927 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 928 * both cgroup-attached and other progs and supports all functionality 929 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 930 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 931 */ 932 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 933 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 934 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 935 BPF_MAP_TYPE_QUEUE, 936 BPF_MAP_TYPE_STACK, 937 BPF_MAP_TYPE_SK_STORAGE, 938 BPF_MAP_TYPE_DEVMAP_HASH, 939 BPF_MAP_TYPE_STRUCT_OPS, 940 BPF_MAP_TYPE_RINGBUF, 941 BPF_MAP_TYPE_INODE_STORAGE, 942 BPF_MAP_TYPE_TASK_STORAGE, 943 BPF_MAP_TYPE_BLOOM_FILTER, 944 BPF_MAP_TYPE_USER_RINGBUF, 945 BPF_MAP_TYPE_CGRP_STORAGE, 946}; 947 948/* Note that tracing related programs such as 949 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 950 * are not subject to a stable API since kernel internal data 951 * structures can change from release to release and may 952 * therefore break existing tracing BPF programs. Tracing BPF 953 * programs correspond to /a/ specific kernel which is to be 954 * analyzed, and not /a/ specific kernel /and/ all future ones. 955 */ 956enum bpf_prog_type { 957 BPF_PROG_TYPE_UNSPEC, 958 BPF_PROG_TYPE_SOCKET_FILTER, 959 BPF_PROG_TYPE_KPROBE, 960 BPF_PROG_TYPE_SCHED_CLS, 961 BPF_PROG_TYPE_SCHED_ACT, 962 BPF_PROG_TYPE_TRACEPOINT, 963 BPF_PROG_TYPE_XDP, 964 BPF_PROG_TYPE_PERF_EVENT, 965 BPF_PROG_TYPE_CGROUP_SKB, 966 BPF_PROG_TYPE_CGROUP_SOCK, 967 BPF_PROG_TYPE_LWT_IN, 968 BPF_PROG_TYPE_LWT_OUT, 969 BPF_PROG_TYPE_LWT_XMIT, 970 BPF_PROG_TYPE_SOCK_OPS, 971 BPF_PROG_TYPE_SK_SKB, 972 BPF_PROG_TYPE_CGROUP_DEVICE, 973 BPF_PROG_TYPE_SK_MSG, 974 BPF_PROG_TYPE_RAW_TRACEPOINT, 975 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 976 BPF_PROG_TYPE_LWT_SEG6LOCAL, 977 BPF_PROG_TYPE_LIRC_MODE2, 978 BPF_PROG_TYPE_SK_REUSEPORT, 979 BPF_PROG_TYPE_FLOW_DISSECTOR, 980 BPF_PROG_TYPE_CGROUP_SYSCTL, 981 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 982 BPF_PROG_TYPE_CGROUP_SOCKOPT, 983 BPF_PROG_TYPE_TRACING, 984 BPF_PROG_TYPE_STRUCT_OPS, 985 BPF_PROG_TYPE_EXT, 986 BPF_PROG_TYPE_LSM, 987 BPF_PROG_TYPE_SK_LOOKUP, 988 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 989 BPF_PROG_TYPE_NETFILTER, 990}; 991 992enum bpf_attach_type { 993 BPF_CGROUP_INET_INGRESS, 994 BPF_CGROUP_INET_EGRESS, 995 BPF_CGROUP_INET_SOCK_CREATE, 996 BPF_CGROUP_SOCK_OPS, 997 BPF_SK_SKB_STREAM_PARSER, 998 BPF_SK_SKB_STREAM_VERDICT, 999 BPF_CGROUP_DEVICE, 1000 BPF_SK_MSG_VERDICT, 1001 BPF_CGROUP_INET4_BIND, 1002 BPF_CGROUP_INET6_BIND, 1003 BPF_CGROUP_INET4_CONNECT, 1004 BPF_CGROUP_INET6_CONNECT, 1005 BPF_CGROUP_INET4_POST_BIND, 1006 BPF_CGROUP_INET6_POST_BIND, 1007 BPF_CGROUP_UDP4_SENDMSG, 1008 BPF_CGROUP_UDP6_SENDMSG, 1009 BPF_LIRC_MODE2, 1010 BPF_FLOW_DISSECTOR, 1011 BPF_CGROUP_SYSCTL, 1012 BPF_CGROUP_UDP4_RECVMSG, 1013 BPF_CGROUP_UDP6_RECVMSG, 1014 BPF_CGROUP_GETSOCKOPT, 1015 BPF_CGROUP_SETSOCKOPT, 1016 BPF_TRACE_RAW_TP, 1017 BPF_TRACE_FENTRY, 1018 BPF_TRACE_FEXIT, 1019 BPF_MODIFY_RETURN, 1020 BPF_LSM_MAC, 1021 BPF_TRACE_ITER, 1022 BPF_CGROUP_INET4_GETPEERNAME, 1023 BPF_CGROUP_INET6_GETPEERNAME, 1024 BPF_CGROUP_INET4_GETSOCKNAME, 1025 BPF_CGROUP_INET6_GETSOCKNAME, 1026 BPF_XDP_DEVMAP, 1027 BPF_CGROUP_INET_SOCK_RELEASE, 1028 BPF_XDP_CPUMAP, 1029 BPF_SK_LOOKUP, 1030 BPF_XDP, 1031 BPF_SK_SKB_VERDICT, 1032 BPF_SK_REUSEPORT_SELECT, 1033 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1034 BPF_PERF_EVENT, 1035 BPF_TRACE_KPROBE_MULTI, 1036 BPF_LSM_CGROUP, 1037 BPF_STRUCT_OPS, 1038 BPF_NETFILTER, 1039 __MAX_BPF_ATTACH_TYPE 1040}; 1041 1042#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1043 1044enum bpf_link_type { 1045 BPF_LINK_TYPE_UNSPEC = 0, 1046 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1047 BPF_LINK_TYPE_TRACING = 2, 1048 BPF_LINK_TYPE_CGROUP = 3, 1049 BPF_LINK_TYPE_ITER = 4, 1050 BPF_LINK_TYPE_NETNS = 5, 1051 BPF_LINK_TYPE_XDP = 6, 1052 BPF_LINK_TYPE_PERF_EVENT = 7, 1053 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1054 BPF_LINK_TYPE_STRUCT_OPS = 9, 1055 BPF_LINK_TYPE_NETFILTER = 10, 1056 1057 MAX_BPF_LINK_TYPE, 1058}; 1059 1060/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1061 * 1062 * NONE(default): No further bpf programs allowed in the subtree. 1063 * 1064 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1065 * the program in this cgroup yields to sub-cgroup program. 1066 * 1067 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1068 * that cgroup program gets run in addition to the program in this cgroup. 1069 * 1070 * Only one program is allowed to be attached to a cgroup with 1071 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1072 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1073 * release old program and attach the new one. Attach flags has to match. 1074 * 1075 * Multiple programs are allowed to be attached to a cgroup with 1076 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1077 * (those that were attached first, run first) 1078 * The programs of sub-cgroup are executed first, then programs of 1079 * this cgroup and then programs of parent cgroup. 1080 * When children program makes decision (like picking TCP CA or sock bind) 1081 * parent program has a chance to override it. 1082 * 1083 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1084 * programs for a cgroup. Though it's possible to replace an old program at 1085 * any position by also specifying BPF_F_REPLACE flag and position itself in 1086 * replace_bpf_fd attribute. Old program at this position will be released. 1087 * 1088 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1089 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1090 * Ex1: 1091 * cgrp1 (MULTI progs A, B) -> 1092 * cgrp2 (OVERRIDE prog C) -> 1093 * cgrp3 (MULTI prog D) -> 1094 * cgrp4 (OVERRIDE prog E) -> 1095 * cgrp5 (NONE prog F) 1096 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1097 * if prog F is detached, the execution is E,D,A,B 1098 * if prog F and D are detached, the execution is E,A,B 1099 * if prog F, E and D are detached, the execution is C,A,B 1100 * 1101 * All eligible programs are executed regardless of return code from 1102 * earlier programs. 1103 */ 1104#define BPF_F_ALLOW_OVERRIDE (1U << 0) 1105#define BPF_F_ALLOW_MULTI (1U << 1) 1106#define BPF_F_REPLACE (1U << 2) 1107 1108/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1109 * verifier will perform strict alignment checking as if the kernel 1110 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1111 * and NET_IP_ALIGN defined to 2. 1112 */ 1113#define BPF_F_STRICT_ALIGNMENT (1U << 0) 1114 1115/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1116 * verifier will allow any alignment whatsoever. On platforms 1117 * with strict alignment requirements for loads ands stores (such 1118 * as sparc and mips) the verifier validates that all loads and 1119 * stores provably follow this requirement. This flag turns that 1120 * checking and enforcement off. 1121 * 1122 * It is mostly used for testing when we want to validate the 1123 * context and memory access aspects of the verifier, but because 1124 * of an unaligned access the alignment check would trigger before 1125 * the one we are interested in. 1126 */ 1127#define BPF_F_ANY_ALIGNMENT (1U << 1) 1128 1129/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1130 * Verifier does sub-register def/use analysis and identifies instructions whose 1131 * def only matters for low 32-bit, high 32-bit is never referenced later 1132 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1133 * that it is safe to ignore clearing high 32-bit for these instructions. This 1134 * saves some back-ends a lot of code-gen. However such optimization is not 1135 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1136 * hence hasn't used verifier's analysis result. But, we really want to have a 1137 * way to be able to verify the correctness of the described optimization on 1138 * x86_64 on which testsuites are frequently exercised. 1139 * 1140 * So, this flag is introduced. Once it is set, verifier will randomize high 1141 * 32-bit for those instructions who has been identified as safe to ignore them. 1142 * Then, if verifier is not doing correct analysis, such randomization will 1143 * regress tests to expose bugs. 1144 */ 1145#define BPF_F_TEST_RND_HI32 (1U << 2) 1146 1147/* The verifier internal test flag. Behavior is undefined */ 1148#define BPF_F_TEST_STATE_FREQ (1U << 3) 1149 1150/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1151 * restrict map and helper usage for such programs. Sleepable BPF programs can 1152 * only be attached to hooks where kernel execution context allows sleeping. 1153 * Such programs are allowed to use helpers that may sleep like 1154 * bpf_copy_from_user(). 1155 */ 1156#define BPF_F_SLEEPABLE (1U << 4) 1157 1158/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1159 * fully support xdp frags. 1160 */ 1161#define BPF_F_XDP_HAS_FRAGS (1U << 5) 1162 1163/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1164 * program becomes device-bound but can access XDP metadata. 1165 */ 1166#define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1167 1168/* link_create.kprobe_multi.flags used in LINK_CREATE command for 1169 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1170 */ 1171#define BPF_F_KPROBE_MULTI_RETURN (1U << 0) 1172 1173/* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1174 * the following extensions: 1175 * 1176 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1177 * insn[0].imm: map fd or fd_idx 1178 * insn[1].imm: 0 1179 * insn[0].off: 0 1180 * insn[1].off: 0 1181 * ldimm64 rewrite: address of map 1182 * verifier type: CONST_PTR_TO_MAP 1183 */ 1184#define BPF_PSEUDO_MAP_FD 1 1185#define BPF_PSEUDO_MAP_IDX 5 1186 1187/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1188 * insn[0].imm: map fd or fd_idx 1189 * insn[1].imm: offset into value 1190 * insn[0].off: 0 1191 * insn[1].off: 0 1192 * ldimm64 rewrite: address of map[0]+offset 1193 * verifier type: PTR_TO_MAP_VALUE 1194 */ 1195#define BPF_PSEUDO_MAP_VALUE 2 1196#define BPF_PSEUDO_MAP_IDX_VALUE 6 1197 1198/* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1199 * insn[0].imm: kernel btd id of VAR 1200 * insn[1].imm: 0 1201 * insn[0].off: 0 1202 * insn[1].off: 0 1203 * ldimm64 rewrite: address of the kernel variable 1204 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1205 * is struct/union. 1206 */ 1207#define BPF_PSEUDO_BTF_ID 3 1208/* insn[0].src_reg: BPF_PSEUDO_FUNC 1209 * insn[0].imm: insn offset to the func 1210 * insn[1].imm: 0 1211 * insn[0].off: 0 1212 * insn[1].off: 0 1213 * ldimm64 rewrite: address of the function 1214 * verifier type: PTR_TO_FUNC. 1215 */ 1216#define BPF_PSEUDO_FUNC 4 1217 1218/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1219 * offset to another bpf function 1220 */ 1221#define BPF_PSEUDO_CALL 1 1222/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1223 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1224 */ 1225#define BPF_PSEUDO_KFUNC_CALL 2 1226 1227/* flags for BPF_MAP_UPDATE_ELEM command */ 1228enum { 1229 BPF_ANY = 0, /* create new element or update existing */ 1230 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1231 BPF_EXIST = 2, /* update existing element */ 1232 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1233}; 1234 1235/* flags for BPF_MAP_CREATE command */ 1236enum { 1237 BPF_F_NO_PREALLOC = (1U << 0), 1238/* Instead of having one common LRU list in the 1239 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1240 * which can scale and perform better. 1241 * Note, the LRU nodes (including free nodes) cannot be moved 1242 * across different LRU lists. 1243 */ 1244 BPF_F_NO_COMMON_LRU = (1U << 1), 1245/* Specify numa node during map creation */ 1246 BPF_F_NUMA_NODE = (1U << 2), 1247 1248/* Flags for accessing BPF object from syscall side. */ 1249 BPF_F_RDONLY = (1U << 3), 1250 BPF_F_WRONLY = (1U << 4), 1251 1252/* Flag for stack_map, store build_id+offset instead of pointer */ 1253 BPF_F_STACK_BUILD_ID = (1U << 5), 1254 1255/* Zero-initialize hash function seed. This should only be used for testing. */ 1256 BPF_F_ZERO_SEED = (1U << 6), 1257 1258/* Flags for accessing BPF object from program side. */ 1259 BPF_F_RDONLY_PROG = (1U << 7), 1260 BPF_F_WRONLY_PROG = (1U << 8), 1261 1262/* Clone map from listener for newly accepted socket */ 1263 BPF_F_CLONE = (1U << 9), 1264 1265/* Enable memory-mapping BPF map */ 1266 BPF_F_MMAPABLE = (1U << 10), 1267 1268/* Share perf_event among processes */ 1269 BPF_F_PRESERVE_ELEMS = (1U << 11), 1270 1271/* Create a map that is suitable to be an inner map with dynamic max entries */ 1272 BPF_F_INNER_MAP = (1U << 12), 1273 1274/* Create a map that will be registered/unregesitered by the backed bpf_link */ 1275 BPF_F_LINK = (1U << 13), 1276 1277/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1278 BPF_F_PATH_FD = (1U << 14), 1279}; 1280 1281/* Flags for BPF_PROG_QUERY. */ 1282 1283/* Query effective (directly attached + inherited from ancestor cgroups) 1284 * programs that will be executed for events within a cgroup. 1285 * attach_flags with this flag are always returned 0. 1286 */ 1287#define BPF_F_QUERY_EFFECTIVE (1U << 0) 1288 1289/* Flags for BPF_PROG_TEST_RUN */ 1290 1291/* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1292#define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1293/* If set, XDP frames will be transmitted after processing */ 1294#define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1295 1296/* type for BPF_ENABLE_STATS */ 1297enum bpf_stats_type { 1298 /* enabled run_time_ns and run_cnt */ 1299 BPF_STATS_RUN_TIME = 0, 1300}; 1301 1302enum bpf_stack_build_id_status { 1303 /* user space need an empty entry to identify end of a trace */ 1304 BPF_STACK_BUILD_ID_EMPTY = 0, 1305 /* with valid build_id and offset */ 1306 BPF_STACK_BUILD_ID_VALID = 1, 1307 /* couldn't get build_id, fallback to ip */ 1308 BPF_STACK_BUILD_ID_IP = 2, 1309}; 1310 1311#define BPF_BUILD_ID_SIZE 20 1312struct bpf_stack_build_id { 1313 __s32 status; 1314 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1315 union { 1316 __u64 offset; 1317 __u64 ip; 1318 }; 1319}; 1320 1321#define BPF_OBJ_NAME_LEN 16U 1322 1323union bpf_attr { 1324 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1325 __u32 map_type; /* one of enum bpf_map_type */ 1326 __u32 key_size; /* size of key in bytes */ 1327 __u32 value_size; /* size of value in bytes */ 1328 __u32 max_entries; /* max number of entries in a map */ 1329 __u32 map_flags; /* BPF_MAP_CREATE related 1330 * flags defined above. 1331 */ 1332 __u32 inner_map_fd; /* fd pointing to the inner map */ 1333 __u32 numa_node; /* numa node (effective only if 1334 * BPF_F_NUMA_NODE is set). 1335 */ 1336 char map_name[BPF_OBJ_NAME_LEN]; 1337 __u32 map_ifindex; /* ifindex of netdev to create on */ 1338 __u32 btf_fd; /* fd pointing to a BTF type data */ 1339 __u32 btf_key_type_id; /* BTF type_id of the key */ 1340 __u32 btf_value_type_id; /* BTF type_id of the value */ 1341 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1342 * struct stored as the 1343 * map value 1344 */ 1345 /* Any per-map-type extra fields 1346 * 1347 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1348 * number of hash functions (if 0, the bloom filter will default 1349 * to using 5 hash functions). 1350 */ 1351 __u64 map_extra; 1352 }; 1353 1354 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1355 __u32 map_fd; 1356 __aligned_u64 key; 1357 union { 1358 __aligned_u64 value; 1359 __aligned_u64 next_key; 1360 }; 1361 __u64 flags; 1362 }; 1363 1364 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1365 __aligned_u64 in_batch; /* start batch, 1366 * NULL to start from beginning 1367 */ 1368 __aligned_u64 out_batch; /* output: next start batch */ 1369 __aligned_u64 keys; 1370 __aligned_u64 values; 1371 __u32 count; /* input/output: 1372 * input: # of key/value 1373 * elements 1374 * output: # of filled elements 1375 */ 1376 __u32 map_fd; 1377 __u64 elem_flags; 1378 __u64 flags; 1379 } batch; 1380 1381 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1382 __u32 prog_type; /* one of enum bpf_prog_type */ 1383 __u32 insn_cnt; 1384 __aligned_u64 insns; 1385 __aligned_u64 license; 1386 __u32 log_level; /* verbosity level of verifier */ 1387 __u32 log_size; /* size of user buffer */ 1388 __aligned_u64 log_buf; /* user supplied buffer */ 1389 __u32 kern_version; /* not used */ 1390 __u32 prog_flags; 1391 char prog_name[BPF_OBJ_NAME_LEN]; 1392 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1393 /* For some prog types expected attach type must be known at 1394 * load time to verify attach type specific parts of prog 1395 * (context accesses, allowed helpers, etc). 1396 */ 1397 __u32 expected_attach_type; 1398 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1399 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1400 __aligned_u64 func_info; /* func info */ 1401 __u32 func_info_cnt; /* number of bpf_func_info records */ 1402 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1403 __aligned_u64 line_info; /* line info */ 1404 __u32 line_info_cnt; /* number of bpf_line_info records */ 1405 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1406 union { 1407 /* valid prog_fd to attach to bpf prog */ 1408 __u32 attach_prog_fd; 1409 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1410 __u32 attach_btf_obj_fd; 1411 }; 1412 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1413 __aligned_u64 fd_array; /* array of FDs */ 1414 __aligned_u64 core_relos; 1415 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1416 /* output: actual total log contents size (including termintaing zero). 1417 * It could be both larger than original log_size (if log was 1418 * truncated), or smaller (if log buffer wasn't filled completely). 1419 */ 1420 __u32 log_true_size; 1421 }; 1422 1423 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1424 __aligned_u64 pathname; 1425 __u32 bpf_fd; 1426 __u32 file_flags; 1427 /* Same as dirfd in openat() syscall; see openat(2) 1428 * manpage for details of path FD and pathname semantics; 1429 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1430 * file_flags field, otherwise it should be set to zero; 1431 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1432 */ 1433 __s32 path_fd; 1434 }; 1435 1436 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1437 __u32 target_fd; /* container object to attach to */ 1438 __u32 attach_bpf_fd; /* eBPF program to attach */ 1439 __u32 attach_type; 1440 __u32 attach_flags; 1441 __u32 replace_bpf_fd; /* previously attached eBPF 1442 * program to replace if 1443 * BPF_F_REPLACE is used 1444 */ 1445 }; 1446 1447 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1448 __u32 prog_fd; 1449 __u32 retval; 1450 __u32 data_size_in; /* input: len of data_in */ 1451 __u32 data_size_out; /* input/output: len of data_out 1452 * returns ENOSPC if data_out 1453 * is too small. 1454 */ 1455 __aligned_u64 data_in; 1456 __aligned_u64 data_out; 1457 __u32 repeat; 1458 __u32 duration; 1459 __u32 ctx_size_in; /* input: len of ctx_in */ 1460 __u32 ctx_size_out; /* input/output: len of ctx_out 1461 * returns ENOSPC if ctx_out 1462 * is too small. 1463 */ 1464 __aligned_u64 ctx_in; 1465 __aligned_u64 ctx_out; 1466 __u32 flags; 1467 __u32 cpu; 1468 __u32 batch_size; 1469 } test; 1470 1471 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1472 union { 1473 __u32 start_id; 1474 __u32 prog_id; 1475 __u32 map_id; 1476 __u32 btf_id; 1477 __u32 link_id; 1478 }; 1479 __u32 next_id; 1480 __u32 open_flags; 1481 }; 1482 1483 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1484 __u32 bpf_fd; 1485 __u32 info_len; 1486 __aligned_u64 info; 1487 } info; 1488 1489 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1490 __u32 target_fd; /* container object to query */ 1491 __u32 attach_type; 1492 __u32 query_flags; 1493 __u32 attach_flags; 1494 __aligned_u64 prog_ids; 1495 __u32 prog_cnt; 1496 /* output: per-program attach_flags. 1497 * not allowed to be set during effective query. 1498 */ 1499 __aligned_u64 prog_attach_flags; 1500 } query; 1501 1502 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1503 __u64 name; 1504 __u32 prog_fd; 1505 } raw_tracepoint; 1506 1507 struct { /* anonymous struct for BPF_BTF_LOAD */ 1508 __aligned_u64 btf; 1509 __aligned_u64 btf_log_buf; 1510 __u32 btf_size; 1511 __u32 btf_log_size; 1512 __u32 btf_log_level; 1513 /* output: actual total log contents size (including termintaing zero). 1514 * It could be both larger than original log_size (if log was 1515 * truncated), or smaller (if log buffer wasn't filled completely). 1516 */ 1517 __u32 btf_log_true_size; 1518 }; 1519 1520 struct { 1521 __u32 pid; /* input: pid */ 1522 __u32 fd; /* input: fd */ 1523 __u32 flags; /* input: flags */ 1524 __u32 buf_len; /* input/output: buf len */ 1525 __aligned_u64 buf; /* input/output: 1526 * tp_name for tracepoint 1527 * symbol for kprobe 1528 * filename for uprobe 1529 */ 1530 __u32 prog_id; /* output: prod_id */ 1531 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1532 __u64 probe_offset; /* output: probe_offset */ 1533 __u64 probe_addr; /* output: probe_addr */ 1534 } task_fd_query; 1535 1536 struct { /* struct used by BPF_LINK_CREATE command */ 1537 union { 1538 __u32 prog_fd; /* eBPF program to attach */ 1539 __u32 map_fd; /* struct_ops to attach */ 1540 }; 1541 union { 1542 __u32 target_fd; /* object to attach to */ 1543 __u32 target_ifindex; /* target ifindex */ 1544 }; 1545 __u32 attach_type; /* attach type */ 1546 __u32 flags; /* extra flags */ 1547 union { 1548 __u32 target_btf_id; /* btf_id of target to attach to */ 1549 struct { 1550 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1551 __u32 iter_info_len; /* iter_info length */ 1552 }; 1553 struct { 1554 /* black box user-provided value passed through 1555 * to BPF program at the execution time and 1556 * accessible through bpf_get_attach_cookie() BPF helper 1557 */ 1558 __u64 bpf_cookie; 1559 } perf_event; 1560 struct { 1561 __u32 flags; 1562 __u32 cnt; 1563 __aligned_u64 syms; 1564 __aligned_u64 addrs; 1565 __aligned_u64 cookies; 1566 } kprobe_multi; 1567 struct { 1568 /* this is overlaid with the target_btf_id above. */ 1569 __u32 target_btf_id; 1570 /* black box user-provided value passed through 1571 * to BPF program at the execution time and 1572 * accessible through bpf_get_attach_cookie() BPF helper 1573 */ 1574 __u64 cookie; 1575 } tracing; 1576 struct { 1577 __u32 pf; 1578 __u32 hooknum; 1579 __s32 priority; 1580 __u32 flags; 1581 } netfilter; 1582 }; 1583 } link_create; 1584 1585 struct { /* struct used by BPF_LINK_UPDATE command */ 1586 __u32 link_fd; /* link fd */ 1587 union { 1588 /* new program fd to update link with */ 1589 __u32 new_prog_fd; 1590 /* new struct_ops map fd to update link with */ 1591 __u32 new_map_fd; 1592 }; 1593 __u32 flags; /* extra flags */ 1594 union { 1595 /* expected link's program fd; is specified only if 1596 * BPF_F_REPLACE flag is set in flags. 1597 */ 1598 __u32 old_prog_fd; 1599 /* expected link's map fd; is specified only 1600 * if BPF_F_REPLACE flag is set. 1601 */ 1602 __u32 old_map_fd; 1603 }; 1604 } link_update; 1605 1606 struct { 1607 __u32 link_fd; 1608 } link_detach; 1609 1610 struct { /* struct used by BPF_ENABLE_STATS command */ 1611 __u32 type; 1612 } enable_stats; 1613 1614 struct { /* struct used by BPF_ITER_CREATE command */ 1615 __u32 link_fd; 1616 __u32 flags; 1617 } iter_create; 1618 1619 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1620 __u32 prog_fd; 1621 __u32 map_fd; 1622 __u32 flags; /* extra flags */ 1623 } prog_bind_map; 1624 1625} __attribute__((aligned(8))); 1626 1627/* The description below is an attempt at providing documentation to eBPF 1628 * developers about the multiple available eBPF helper functions. It can be 1629 * parsed and used to produce a manual page. The workflow is the following, 1630 * and requires the rst2man utility: 1631 * 1632 * $ ./scripts/bpf_doc.py \ 1633 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1634 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1635 * $ man /tmp/bpf-helpers.7 1636 * 1637 * Note that in order to produce this external documentation, some RST 1638 * formatting is used in the descriptions to get "bold" and "italics" in 1639 * manual pages. Also note that the few trailing white spaces are 1640 * intentional, removing them would break paragraphs for rst2man. 1641 * 1642 * Start of BPF helper function descriptions: 1643 * 1644 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1645 * Description 1646 * Perform a lookup in *map* for an entry associated to *key*. 1647 * Return 1648 * Map value associated to *key*, or **NULL** if no entry was 1649 * found. 1650 * 1651 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1652 * Description 1653 * Add or update the value of the entry associated to *key* in 1654 * *map* with *value*. *flags* is one of: 1655 * 1656 * **BPF_NOEXIST** 1657 * The entry for *key* must not exist in the map. 1658 * **BPF_EXIST** 1659 * The entry for *key* must already exist in the map. 1660 * **BPF_ANY** 1661 * No condition on the existence of the entry for *key*. 1662 * 1663 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1664 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1665 * elements always exist), the helper would return an error. 1666 * Return 1667 * 0 on success, or a negative error in case of failure. 1668 * 1669 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1670 * Description 1671 * Delete entry with *key* from *map*. 1672 * Return 1673 * 0 on success, or a negative error in case of failure. 1674 * 1675 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1676 * Description 1677 * For tracing programs, safely attempt to read *size* bytes from 1678 * kernel space address *unsafe_ptr* and store the data in *dst*. 1679 * 1680 * Generally, use **bpf_probe_read_user**\ () or 1681 * **bpf_probe_read_kernel**\ () instead. 1682 * Return 1683 * 0 on success, or a negative error in case of failure. 1684 * 1685 * u64 bpf_ktime_get_ns(void) 1686 * Description 1687 * Return the time elapsed since system boot, in nanoseconds. 1688 * Does not include time the system was suspended. 1689 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1690 * Return 1691 * Current *ktime*. 1692 * 1693 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1694 * Description 1695 * This helper is a "printk()-like" facility for debugging. It 1696 * prints a message defined by format *fmt* (of size *fmt_size*) 1697 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1698 * available. It can take up to three additional **u64** 1699 * arguments (as an eBPF helpers, the total number of arguments is 1700 * limited to five). 1701 * 1702 * Each time the helper is called, it appends a line to the trace. 1703 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1704 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1705 * The format of the trace is customizable, and the exact output 1706 * one will get depends on the options set in 1707 * *\/sys/kernel/tracing/trace_options* (see also the 1708 * *README* file under the same directory). However, it usually 1709 * defaults to something like: 1710 * 1711 * :: 1712 * 1713 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1714 * 1715 * In the above: 1716 * 1717 * * ``telnet`` is the name of the current task. 1718 * * ``470`` is the PID of the current task. 1719 * * ``001`` is the CPU number on which the task is 1720 * running. 1721 * * In ``.N..``, each character refers to a set of 1722 * options (whether irqs are enabled, scheduling 1723 * options, whether hard/softirqs are running, level of 1724 * preempt_disabled respectively). **N** means that 1725 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1726 * are set. 1727 * * ``419421.045894`` is a timestamp. 1728 * * ``0x00000001`` is a fake value used by BPF for the 1729 * instruction pointer register. 1730 * * ``<formatted msg>`` is the message formatted with 1731 * *fmt*. 1732 * 1733 * The conversion specifiers supported by *fmt* are similar, but 1734 * more limited than for printk(). They are **%d**, **%i**, 1735 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1736 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1737 * of field, padding with zeroes, etc.) is available, and the 1738 * helper will return **-EINVAL** (but print nothing) if it 1739 * encounters an unknown specifier. 1740 * 1741 * Also, note that **bpf_trace_printk**\ () is slow, and should 1742 * only be used for debugging purposes. For this reason, a notice 1743 * block (spanning several lines) is printed to kernel logs and 1744 * states that the helper should not be used "for production use" 1745 * the first time this helper is used (or more precisely, when 1746 * **trace_printk**\ () buffers are allocated). For passing values 1747 * to user space, perf events should be preferred. 1748 * Return 1749 * The number of bytes written to the buffer, or a negative error 1750 * in case of failure. 1751 * 1752 * u32 bpf_get_prandom_u32(void) 1753 * Description 1754 * Get a pseudo-random number. 1755 * 1756 * From a security point of view, this helper uses its own 1757 * pseudo-random internal state, and cannot be used to infer the 1758 * seed of other random functions in the kernel. However, it is 1759 * essential to note that the generator used by the helper is not 1760 * cryptographically secure. 1761 * Return 1762 * A random 32-bit unsigned value. 1763 * 1764 * u32 bpf_get_smp_processor_id(void) 1765 * Description 1766 * Get the SMP (symmetric multiprocessing) processor id. Note that 1767 * all programs run with migration disabled, which means that the 1768 * SMP processor id is stable during all the execution of the 1769 * program. 1770 * Return 1771 * The SMP id of the processor running the program. 1772 * 1773 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1774 * Description 1775 * Store *len* bytes from address *from* into the packet 1776 * associated to *skb*, at *offset*. *flags* are a combination of 1777 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1778 * checksum for the packet after storing the bytes) and 1779 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1780 * **->swhash** and *skb*\ **->l4hash** to 0). 1781 * 1782 * A call to this helper is susceptible to change the underlying 1783 * packet buffer. Therefore, at load time, all checks on pointers 1784 * previously done by the verifier are invalidated and must be 1785 * performed again, if the helper is used in combination with 1786 * direct packet access. 1787 * Return 1788 * 0 on success, or a negative error in case of failure. 1789 * 1790 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1791 * Description 1792 * Recompute the layer 3 (e.g. IP) checksum for the packet 1793 * associated to *skb*. Computation is incremental, so the helper 1794 * must know the former value of the header field that was 1795 * modified (*from*), the new value of this field (*to*), and the 1796 * number of bytes (2 or 4) for this field, stored in *size*. 1797 * Alternatively, it is possible to store the difference between 1798 * the previous and the new values of the header field in *to*, by 1799 * setting *from* and *size* to 0. For both methods, *offset* 1800 * indicates the location of the IP checksum within the packet. 1801 * 1802 * This helper works in combination with **bpf_csum_diff**\ (), 1803 * which does not update the checksum in-place, but offers more 1804 * flexibility and can handle sizes larger than 2 or 4 for the 1805 * checksum to update. 1806 * 1807 * A call to this helper is susceptible to change the underlying 1808 * packet buffer. Therefore, at load time, all checks on pointers 1809 * previously done by the verifier are invalidated and must be 1810 * performed again, if the helper is used in combination with 1811 * direct packet access. 1812 * Return 1813 * 0 on success, or a negative error in case of failure. 1814 * 1815 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1816 * Description 1817 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1818 * packet associated to *skb*. Computation is incremental, so the 1819 * helper must know the former value of the header field that was 1820 * modified (*from*), the new value of this field (*to*), and the 1821 * number of bytes (2 or 4) for this field, stored on the lowest 1822 * four bits of *flags*. Alternatively, it is possible to store 1823 * the difference between the previous and the new values of the 1824 * header field in *to*, by setting *from* and the four lowest 1825 * bits of *flags* to 0. For both methods, *offset* indicates the 1826 * location of the IP checksum within the packet. In addition to 1827 * the size of the field, *flags* can be added (bitwise OR) actual 1828 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1829 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1830 * for updates resulting in a null checksum the value is set to 1831 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1832 * the checksum is to be computed against a pseudo-header. 1833 * 1834 * This helper works in combination with **bpf_csum_diff**\ (), 1835 * which does not update the checksum in-place, but offers more 1836 * flexibility and can handle sizes larger than 2 or 4 for the 1837 * checksum to update. 1838 * 1839 * A call to this helper is susceptible to change the underlying 1840 * packet buffer. Therefore, at load time, all checks on pointers 1841 * previously done by the verifier are invalidated and must be 1842 * performed again, if the helper is used in combination with 1843 * direct packet access. 1844 * Return 1845 * 0 on success, or a negative error in case of failure. 1846 * 1847 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1848 * Description 1849 * This special helper is used to trigger a "tail call", or in 1850 * other words, to jump into another eBPF program. The same stack 1851 * frame is used (but values on stack and in registers for the 1852 * caller are not accessible to the callee). This mechanism allows 1853 * for program chaining, either for raising the maximum number of 1854 * available eBPF instructions, or to execute given programs in 1855 * conditional blocks. For security reasons, there is an upper 1856 * limit to the number of successive tail calls that can be 1857 * performed. 1858 * 1859 * Upon call of this helper, the program attempts to jump into a 1860 * program referenced at index *index* in *prog_array_map*, a 1861 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1862 * *ctx*, a pointer to the context. 1863 * 1864 * If the call succeeds, the kernel immediately runs the first 1865 * instruction of the new program. This is not a function call, 1866 * and it never returns to the previous program. If the call 1867 * fails, then the helper has no effect, and the caller continues 1868 * to run its subsequent instructions. A call can fail if the 1869 * destination program for the jump does not exist (i.e. *index* 1870 * is superior to the number of entries in *prog_array_map*), or 1871 * if the maximum number of tail calls has been reached for this 1872 * chain of programs. This limit is defined in the kernel by the 1873 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1874 * which is currently set to 33. 1875 * Return 1876 * 0 on success, or a negative error in case of failure. 1877 * 1878 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1879 * Description 1880 * Clone and redirect the packet associated to *skb* to another 1881 * net device of index *ifindex*. Both ingress and egress 1882 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1883 * value in *flags* is used to make the distinction (ingress path 1884 * is selected if the flag is present, egress path otherwise). 1885 * This is the only flag supported for now. 1886 * 1887 * In comparison with **bpf_redirect**\ () helper, 1888 * **bpf_clone_redirect**\ () has the associated cost of 1889 * duplicating the packet buffer, but this can be executed out of 1890 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1891 * efficient, but it is handled through an action code where the 1892 * redirection happens only after the eBPF program has returned. 1893 * 1894 * A call to this helper is susceptible to change the underlying 1895 * packet buffer. Therefore, at load time, all checks on pointers 1896 * previously done by the verifier are invalidated and must be 1897 * performed again, if the helper is used in combination with 1898 * direct packet access. 1899 * Return 1900 * 0 on success, or a negative error in case of failure. 1901 * 1902 * u64 bpf_get_current_pid_tgid(void) 1903 * Description 1904 * Get the current pid and tgid. 1905 * Return 1906 * A 64-bit integer containing the current tgid and pid, and 1907 * created as such: 1908 * *current_task*\ **->tgid << 32 \|** 1909 * *current_task*\ **->pid**. 1910 * 1911 * u64 bpf_get_current_uid_gid(void) 1912 * Description 1913 * Get the current uid and gid. 1914 * Return 1915 * A 64-bit integer containing the current GID and UID, and 1916 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1917 * 1918 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1919 * Description 1920 * Copy the **comm** attribute of the current task into *buf* of 1921 * *size_of_buf*. The **comm** attribute contains the name of 1922 * the executable (excluding the path) for the current task. The 1923 * *size_of_buf* must be strictly positive. On success, the 1924 * helper makes sure that the *buf* is NUL-terminated. On failure, 1925 * it is filled with zeroes. 1926 * Return 1927 * 0 on success, or a negative error in case of failure. 1928 * 1929 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1930 * Description 1931 * Retrieve the classid for the current task, i.e. for the net_cls 1932 * cgroup to which *skb* belongs. 1933 * 1934 * This helper can be used on TC egress path, but not on ingress. 1935 * 1936 * The net_cls cgroup provides an interface to tag network packets 1937 * based on a user-provided identifier for all traffic coming from 1938 * the tasks belonging to the related cgroup. See also the related 1939 * kernel documentation, available from the Linux sources in file 1940 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1941 * 1942 * The Linux kernel has two versions for cgroups: there are 1943 * cgroups v1 and cgroups v2. Both are available to users, who can 1944 * use a mixture of them, but note that the net_cls cgroup is for 1945 * cgroup v1 only. This makes it incompatible with BPF programs 1946 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1947 * only hold data for one version of cgroups at a time). 1948 * 1949 * This helper is only available is the kernel was compiled with 1950 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1951 * "**y**" or to "**m**". 1952 * Return 1953 * The classid, or 0 for the default unconfigured classid. 1954 * 1955 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1956 * Description 1957 * Push a *vlan_tci* (VLAN tag control information) of protocol 1958 * *vlan_proto* to the packet associated to *skb*, then update 1959 * the checksum. Note that if *vlan_proto* is different from 1960 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1961 * be **ETH_P_8021Q**. 1962 * 1963 * A call to this helper is susceptible to change the underlying 1964 * packet buffer. Therefore, at load time, all checks on pointers 1965 * previously done by the verifier are invalidated and must be 1966 * performed again, if the helper is used in combination with 1967 * direct packet access. 1968 * Return 1969 * 0 on success, or a negative error in case of failure. 1970 * 1971 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1972 * Description 1973 * Pop a VLAN header from the packet associated to *skb*. 1974 * 1975 * A call to this helper is susceptible to change the underlying 1976 * packet buffer. Therefore, at load time, all checks on pointers 1977 * previously done by the verifier are invalidated and must be 1978 * performed again, if the helper is used in combination with 1979 * direct packet access. 1980 * Return 1981 * 0 on success, or a negative error in case of failure. 1982 * 1983 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1984 * Description 1985 * Get tunnel metadata. This helper takes a pointer *key* to an 1986 * empty **struct bpf_tunnel_key** of **size**, that will be 1987 * filled with tunnel metadata for the packet associated to *skb*. 1988 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1989 * indicates that the tunnel is based on IPv6 protocol instead of 1990 * IPv4. 1991 * 1992 * The **struct bpf_tunnel_key** is an object that generalizes the 1993 * principal parameters used by various tunneling protocols into a 1994 * single struct. This way, it can be used to easily make a 1995 * decision based on the contents of the encapsulation header, 1996 * "summarized" in this struct. In particular, it holds the IP 1997 * address of the remote end (IPv4 or IPv6, depending on the case) 1998 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1999 * this struct exposes the *key*\ **->tunnel_id**, which is 2000 * generally mapped to a VNI (Virtual Network Identifier), making 2001 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2002 * () helper. 2003 * 2004 * Let's imagine that the following code is part of a program 2005 * attached to the TC ingress interface, on one end of a GRE 2006 * tunnel, and is supposed to filter out all messages coming from 2007 * remote ends with IPv4 address other than 10.0.0.1: 2008 * 2009 * :: 2010 * 2011 * int ret; 2012 * struct bpf_tunnel_key key = {}; 2013 * 2014 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2015 * if (ret < 0) 2016 * return TC_ACT_SHOT; // drop packet 2017 * 2018 * if (key.remote_ipv4 != 0x0a000001) 2019 * return TC_ACT_SHOT; // drop packet 2020 * 2021 * return TC_ACT_OK; // accept packet 2022 * 2023 * This interface can also be used with all encapsulation devices 2024 * that can operate in "collect metadata" mode: instead of having 2025 * one network device per specific configuration, the "collect 2026 * metadata" mode only requires a single device where the 2027 * configuration can be extracted from this helper. 2028 * 2029 * This can be used together with various tunnels such as VXLan, 2030 * Geneve, GRE or IP in IP (IPIP). 2031 * Return 2032 * 0 on success, or a negative error in case of failure. 2033 * 2034 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2035 * Description 2036 * Populate tunnel metadata for packet associated to *skb.* The 2037 * tunnel metadata is set to the contents of *key*, of *size*. The 2038 * *flags* can be set to a combination of the following values: 2039 * 2040 * **BPF_F_TUNINFO_IPV6** 2041 * Indicate that the tunnel is based on IPv6 protocol 2042 * instead of IPv4. 2043 * **BPF_F_ZERO_CSUM_TX** 2044 * For IPv4 packets, add a flag to tunnel metadata 2045 * indicating that checksum computation should be skipped 2046 * and checksum set to zeroes. 2047 * **BPF_F_DONT_FRAGMENT** 2048 * Add a flag to tunnel metadata indicating that the 2049 * packet should not be fragmented. 2050 * **BPF_F_SEQ_NUMBER** 2051 * Add a flag to tunnel metadata indicating that a 2052 * sequence number should be added to tunnel header before 2053 * sending the packet. This flag was added for GRE 2054 * encapsulation, but might be used with other protocols 2055 * as well in the future. 2056 * **BPF_F_NO_TUNNEL_KEY** 2057 * Add a flag to tunnel metadata indicating that no tunnel 2058 * key should be set in the resulting tunnel header. 2059 * 2060 * Here is a typical usage on the transmit path: 2061 * 2062 * :: 2063 * 2064 * struct bpf_tunnel_key key; 2065 * populate key ... 2066 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2067 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2068 * 2069 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2070 * helper for additional information. 2071 * Return 2072 * 0 on success, or a negative error in case of failure. 2073 * 2074 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2075 * Description 2076 * Read the value of a perf event counter. This helper relies on a 2077 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2078 * the perf event counter is selected when *map* is updated with 2079 * perf event file descriptors. The *map* is an array whose size 2080 * is the number of available CPUs, and each cell contains a value 2081 * relative to one CPU. The value to retrieve is indicated by 2082 * *flags*, that contains the index of the CPU to look up, masked 2083 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2084 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2085 * current CPU should be retrieved. 2086 * 2087 * Note that before Linux 4.13, only hardware perf event can be 2088 * retrieved. 2089 * 2090 * Also, be aware that the newer helper 2091 * **bpf_perf_event_read_value**\ () is recommended over 2092 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2093 * quirks where error and counter value are used as a return code 2094 * (which is wrong to do since ranges may overlap). This issue is 2095 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2096 * time provides more features over the **bpf_perf_event_read**\ 2097 * () interface. Please refer to the description of 2098 * **bpf_perf_event_read_value**\ () for details. 2099 * Return 2100 * The value of the perf event counter read from the map, or a 2101 * negative error code in case of failure. 2102 * 2103 * long bpf_redirect(u32 ifindex, u64 flags) 2104 * Description 2105 * Redirect the packet to another net device of index *ifindex*. 2106 * This helper is somewhat similar to **bpf_clone_redirect**\ 2107 * (), except that the packet is not cloned, which provides 2108 * increased performance. 2109 * 2110 * Except for XDP, both ingress and egress interfaces can be used 2111 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2112 * to make the distinction (ingress path is selected if the flag 2113 * is present, egress path otherwise). Currently, XDP only 2114 * supports redirection to the egress interface, and accepts no 2115 * flag at all. 2116 * 2117 * The same effect can also be attained with the more generic 2118 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2119 * redirect target instead of providing it directly to the helper. 2120 * Return 2121 * For XDP, the helper returns **XDP_REDIRECT** on success or 2122 * **XDP_ABORTED** on error. For other program types, the values 2123 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2124 * error. 2125 * 2126 * u32 bpf_get_route_realm(struct sk_buff *skb) 2127 * Description 2128 * Retrieve the realm or the route, that is to say the 2129 * **tclassid** field of the destination for the *skb*. The 2130 * identifier retrieved is a user-provided tag, similar to the 2131 * one used with the net_cls cgroup (see description for 2132 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2133 * held by a route (a destination entry), not by a task. 2134 * 2135 * Retrieving this identifier works with the clsact TC egress hook 2136 * (see also **tc-bpf(8)**), or alternatively on conventional 2137 * classful egress qdiscs, but not on TC ingress path. In case of 2138 * clsact TC egress hook, this has the advantage that, internally, 2139 * the destination entry has not been dropped yet in the transmit 2140 * path. Therefore, the destination entry does not need to be 2141 * artificially held via **netif_keep_dst**\ () for a classful 2142 * qdisc until the *skb* is freed. 2143 * 2144 * This helper is available only if the kernel was compiled with 2145 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2146 * Return 2147 * The realm of the route for the packet associated to *skb*, or 0 2148 * if none was found. 2149 * 2150 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2151 * Description 2152 * Write raw *data* blob into a special BPF perf event held by 2153 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2154 * event must have the following attributes: **PERF_SAMPLE_RAW** 2155 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2156 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2157 * 2158 * The *flags* are used to indicate the index in *map* for which 2159 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2160 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2161 * to indicate that the index of the current CPU core should be 2162 * used. 2163 * 2164 * The value to write, of *size*, is passed through eBPF stack and 2165 * pointed by *data*. 2166 * 2167 * The context of the program *ctx* needs also be passed to the 2168 * helper. 2169 * 2170 * On user space, a program willing to read the values needs to 2171 * call **perf_event_open**\ () on the perf event (either for 2172 * one or for all CPUs) and to store the file descriptor into the 2173 * *map*. This must be done before the eBPF program can send data 2174 * into it. An example is available in file 2175 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2176 * tree (the eBPF program counterpart is in 2177 * *samples/bpf/trace_output_kern.c*). 2178 * 2179 * **bpf_perf_event_output**\ () achieves better performance 2180 * than **bpf_trace_printk**\ () for sharing data with user 2181 * space, and is much better suitable for streaming data from eBPF 2182 * programs. 2183 * 2184 * Note that this helper is not restricted to tracing use cases 2185 * and can be used with programs attached to TC or XDP as well, 2186 * where it allows for passing data to user space listeners. Data 2187 * can be: 2188 * 2189 * * Only custom structs, 2190 * * Only the packet payload, or 2191 * * A combination of both. 2192 * Return 2193 * 0 on success, or a negative error in case of failure. 2194 * 2195 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2196 * Description 2197 * This helper was provided as an easy way to load data from a 2198 * packet. It can be used to load *len* bytes from *offset* from 2199 * the packet associated to *skb*, into the buffer pointed by 2200 * *to*. 2201 * 2202 * Since Linux 4.7, usage of this helper has mostly been replaced 2203 * by "direct packet access", enabling packet data to be 2204 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2205 * pointing respectively to the first byte of packet data and to 2206 * the byte after the last byte of packet data. However, it 2207 * remains useful if one wishes to read large quantities of data 2208 * at once from a packet into the eBPF stack. 2209 * Return 2210 * 0 on success, or a negative error in case of failure. 2211 * 2212 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2213 * Description 2214 * Walk a user or a kernel stack and return its id. To achieve 2215 * this, the helper needs *ctx*, which is a pointer to the context 2216 * on which the tracing program is executed, and a pointer to a 2217 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2218 * 2219 * The last argument, *flags*, holds the number of stack frames to 2220 * skip (from 0 to 255), masked with 2221 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2222 * a combination of the following flags: 2223 * 2224 * **BPF_F_USER_STACK** 2225 * Collect a user space stack instead of a kernel stack. 2226 * **BPF_F_FAST_STACK_CMP** 2227 * Compare stacks by hash only. 2228 * **BPF_F_REUSE_STACKID** 2229 * If two different stacks hash into the same *stackid*, 2230 * discard the old one. 2231 * 2232 * The stack id retrieved is a 32 bit long integer handle which 2233 * can be further combined with other data (including other stack 2234 * ids) and used as a key into maps. This can be useful for 2235 * generating a variety of graphs (such as flame graphs or off-cpu 2236 * graphs). 2237 * 2238 * For walking a stack, this helper is an improvement over 2239 * **bpf_probe_read**\ (), which can be used with unrolled loops 2240 * but is not efficient and consumes a lot of eBPF instructions. 2241 * Instead, **bpf_get_stackid**\ () can collect up to 2242 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2243 * this limit can be controlled with the **sysctl** program, and 2244 * that it should be manually increased in order to profile long 2245 * user stacks (such as stacks for Java programs). To do so, use: 2246 * 2247 * :: 2248 * 2249 * # sysctl kernel.perf_event_max_stack=<new value> 2250 * Return 2251 * The positive or null stack id on success, or a negative error 2252 * in case of failure. 2253 * 2254 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2255 * Description 2256 * Compute a checksum difference, from the raw buffer pointed by 2257 * *from*, of length *from_size* (that must be a multiple of 4), 2258 * towards the raw buffer pointed by *to*, of size *to_size* 2259 * (same remark). An optional *seed* can be added to the value 2260 * (this can be cascaded, the seed may come from a previous call 2261 * to the helper). 2262 * 2263 * This is flexible enough to be used in several ways: 2264 * 2265 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2266 * checksum, it can be used when pushing new data. 2267 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2268 * checksum, it can be used when removing data from a packet. 2269 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2270 * can be used to compute a diff. Note that *from_size* and 2271 * *to_size* do not need to be equal. 2272 * 2273 * This helper can be used in combination with 2274 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2275 * which one can feed in the difference computed with 2276 * **bpf_csum_diff**\ (). 2277 * Return 2278 * The checksum result, or a negative error code in case of 2279 * failure. 2280 * 2281 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2282 * Description 2283 * Retrieve tunnel options metadata for the packet associated to 2284 * *skb*, and store the raw tunnel option data to the buffer *opt* 2285 * of *size*. 2286 * 2287 * This helper can be used with encapsulation devices that can 2288 * operate in "collect metadata" mode (please refer to the related 2289 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2290 * more details). A particular example where this can be used is 2291 * in combination with the Geneve encapsulation protocol, where it 2292 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2293 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2294 * the eBPF program. This allows for full customization of these 2295 * headers. 2296 * Return 2297 * The size of the option data retrieved. 2298 * 2299 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2300 * Description 2301 * Set tunnel options metadata for the packet associated to *skb* 2302 * to the option data contained in the raw buffer *opt* of *size*. 2303 * 2304 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2305 * helper for additional information. 2306 * Return 2307 * 0 on success, or a negative error in case of failure. 2308 * 2309 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2310 * Description 2311 * Change the protocol of the *skb* to *proto*. Currently 2312 * supported are transition from IPv4 to IPv6, and from IPv6 to 2313 * IPv4. The helper takes care of the groundwork for the 2314 * transition, including resizing the socket buffer. The eBPF 2315 * program is expected to fill the new headers, if any, via 2316 * **skb_store_bytes**\ () and to recompute the checksums with 2317 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2318 * (). The main case for this helper is to perform NAT64 2319 * operations out of an eBPF program. 2320 * 2321 * Internally, the GSO type is marked as dodgy so that headers are 2322 * checked and segments are recalculated by the GSO/GRO engine. 2323 * The size for GSO target is adapted as well. 2324 * 2325 * All values for *flags* are reserved for future usage, and must 2326 * be left at zero. 2327 * 2328 * A call to this helper is susceptible to change the underlying 2329 * packet buffer. Therefore, at load time, all checks on pointers 2330 * previously done by the verifier are invalidated and must be 2331 * performed again, if the helper is used in combination with 2332 * direct packet access. 2333 * Return 2334 * 0 on success, or a negative error in case of failure. 2335 * 2336 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2337 * Description 2338 * Change the packet type for the packet associated to *skb*. This 2339 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2340 * the eBPF program does not have a write access to *skb*\ 2341 * **->pkt_type** beside this helper. Using a helper here allows 2342 * for graceful handling of errors. 2343 * 2344 * The major use case is to change incoming *skb*s to 2345 * **PACKET_HOST** in a programmatic way instead of having to 2346 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2347 * example. 2348 * 2349 * Note that *type* only allows certain values. At this time, they 2350 * are: 2351 * 2352 * **PACKET_HOST** 2353 * Packet is for us. 2354 * **PACKET_BROADCAST** 2355 * Send packet to all. 2356 * **PACKET_MULTICAST** 2357 * Send packet to group. 2358 * **PACKET_OTHERHOST** 2359 * Send packet to someone else. 2360 * Return 2361 * 0 on success, or a negative error in case of failure. 2362 * 2363 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2364 * Description 2365 * Check whether *skb* is a descendant of the cgroup2 held by 2366 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2367 * Return 2368 * The return value depends on the result of the test, and can be: 2369 * 2370 * * 0, if the *skb* failed the cgroup2 descendant test. 2371 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2372 * * A negative error code, if an error occurred. 2373 * 2374 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2375 * Description 2376 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2377 * not set, in particular if the hash was cleared due to mangling, 2378 * recompute this hash. Later accesses to the hash can be done 2379 * directly with *skb*\ **->hash**. 2380 * 2381 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2382 * prototype with **bpf_skb_change_proto**\ (), or calling 2383 * **bpf_skb_store_bytes**\ () with the 2384 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2385 * the hash and to trigger a new computation for the next call to 2386 * **bpf_get_hash_recalc**\ (). 2387 * Return 2388 * The 32-bit hash. 2389 * 2390 * u64 bpf_get_current_task(void) 2391 * Description 2392 * Get the current task. 2393 * Return 2394 * A pointer to the current task struct. 2395 * 2396 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2397 * Description 2398 * Attempt in a safe way to write *len* bytes from the buffer 2399 * *src* to *dst* in memory. It only works for threads that are in 2400 * user context, and *dst* must be a valid user space address. 2401 * 2402 * This helper should not be used to implement any kind of 2403 * security mechanism because of TOC-TOU attacks, but rather to 2404 * debug, divert, and manipulate execution of semi-cooperative 2405 * processes. 2406 * 2407 * Keep in mind that this feature is meant for experiments, and it 2408 * has a risk of crashing the system and running programs. 2409 * Therefore, when an eBPF program using this helper is attached, 2410 * a warning including PID and process name is printed to kernel 2411 * logs. 2412 * Return 2413 * 0 on success, or a negative error in case of failure. 2414 * 2415 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2416 * Description 2417 * Check whether the probe is being run is the context of a given 2418 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2419 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2420 * Return 2421 * The return value depends on the result of the test, and can be: 2422 * 2423 * * 1, if current task belongs to the cgroup2. 2424 * * 0, if current task does not belong to the cgroup2. 2425 * * A negative error code, if an error occurred. 2426 * 2427 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2428 * Description 2429 * Resize (trim or grow) the packet associated to *skb* to the 2430 * new *len*. The *flags* are reserved for future usage, and must 2431 * be left at zero. 2432 * 2433 * The basic idea is that the helper performs the needed work to 2434 * change the size of the packet, then the eBPF program rewrites 2435 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2436 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2437 * and others. This helper is a slow path utility intended for 2438 * replies with control messages. And because it is targeted for 2439 * slow path, the helper itself can afford to be slow: it 2440 * implicitly linearizes, unclones and drops offloads from the 2441 * *skb*. 2442 * 2443 * A call to this helper is susceptible to change the underlying 2444 * packet buffer. Therefore, at load time, all checks on pointers 2445 * previously done by the verifier are invalidated and must be 2446 * performed again, if the helper is used in combination with 2447 * direct packet access. 2448 * Return 2449 * 0 on success, or a negative error in case of failure. 2450 * 2451 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2452 * Description 2453 * Pull in non-linear data in case the *skb* is non-linear and not 2454 * all of *len* are part of the linear section. Make *len* bytes 2455 * from *skb* readable and writable. If a zero value is passed for 2456 * *len*, then all bytes in the linear part of *skb* will be made 2457 * readable and writable. 2458 * 2459 * This helper is only needed for reading and writing with direct 2460 * packet access. 2461 * 2462 * For direct packet access, testing that offsets to access 2463 * are within packet boundaries (test on *skb*\ **->data_end**) is 2464 * susceptible to fail if offsets are invalid, or if the requested 2465 * data is in non-linear parts of the *skb*. On failure the 2466 * program can just bail out, or in the case of a non-linear 2467 * buffer, use a helper to make the data available. The 2468 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2469 * the data. Another one consists in using **bpf_skb_pull_data** 2470 * to pull in once the non-linear parts, then retesting and 2471 * eventually access the data. 2472 * 2473 * At the same time, this also makes sure the *skb* is uncloned, 2474 * which is a necessary condition for direct write. As this needs 2475 * to be an invariant for the write part only, the verifier 2476 * detects writes and adds a prologue that is calling 2477 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2478 * the very beginning in case it is indeed cloned. 2479 * 2480 * A call to this helper is susceptible to change the underlying 2481 * packet buffer. Therefore, at load time, all checks on pointers 2482 * previously done by the verifier are invalidated and must be 2483 * performed again, if the helper is used in combination with 2484 * direct packet access. 2485 * Return 2486 * 0 on success, or a negative error in case of failure. 2487 * 2488 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2489 * Description 2490 * Add the checksum *csum* into *skb*\ **->csum** in case the 2491 * driver has supplied a checksum for the entire packet into that 2492 * field. Return an error otherwise. This helper is intended to be 2493 * used in combination with **bpf_csum_diff**\ (), in particular 2494 * when the checksum needs to be updated after data has been 2495 * written into the packet through direct packet access. 2496 * Return 2497 * The checksum on success, or a negative error code in case of 2498 * failure. 2499 * 2500 * void bpf_set_hash_invalid(struct sk_buff *skb) 2501 * Description 2502 * Invalidate the current *skb*\ **->hash**. It can be used after 2503 * mangling on headers through direct packet access, in order to 2504 * indicate that the hash is outdated and to trigger a 2505 * recalculation the next time the kernel tries to access this 2506 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2507 * Return 2508 * void. 2509 * 2510 * long bpf_get_numa_node_id(void) 2511 * Description 2512 * Return the id of the current NUMA node. The primary use case 2513 * for this helper is the selection of sockets for the local NUMA 2514 * node, when the program is attached to sockets using the 2515 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2516 * but the helper is also available to other eBPF program types, 2517 * similarly to **bpf_get_smp_processor_id**\ (). 2518 * Return 2519 * The id of current NUMA node. 2520 * 2521 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2522 * Description 2523 * Grows headroom of packet associated to *skb* and adjusts the 2524 * offset of the MAC header accordingly, adding *len* bytes of 2525 * space. It automatically extends and reallocates memory as 2526 * required. 2527 * 2528 * This helper can be used on a layer 3 *skb* to push a MAC header 2529 * for redirection into a layer 2 device. 2530 * 2531 * All values for *flags* are reserved for future usage, and must 2532 * be left at zero. 2533 * 2534 * A call to this helper is susceptible to change the underlying 2535 * packet buffer. Therefore, at load time, all checks on pointers 2536 * previously done by the verifier are invalidated and must be 2537 * performed again, if the helper is used in combination with 2538 * direct packet access. 2539 * Return 2540 * 0 on success, or a negative error in case of failure. 2541 * 2542 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2543 * Description 2544 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2545 * it is possible to use a negative value for *delta*. This helper 2546 * can be used to prepare the packet for pushing or popping 2547 * headers. 2548 * 2549 * A call to this helper is susceptible to change the underlying 2550 * packet buffer. Therefore, at load time, all checks on pointers 2551 * previously done by the verifier are invalidated and must be 2552 * performed again, if the helper is used in combination with 2553 * direct packet access. 2554 * Return 2555 * 0 on success, or a negative error in case of failure. 2556 * 2557 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2558 * Description 2559 * Copy a NUL terminated string from an unsafe kernel address 2560 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2561 * more details. 2562 * 2563 * Generally, use **bpf_probe_read_user_str**\ () or 2564 * **bpf_probe_read_kernel_str**\ () instead. 2565 * Return 2566 * On success, the strictly positive length of the string, 2567 * including the trailing NUL character. On error, a negative 2568 * value. 2569 * 2570 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2571 * Description 2572 * If the **struct sk_buff** pointed by *skb* has a known socket, 2573 * retrieve the cookie (generated by the kernel) of this socket. 2574 * If no cookie has been set yet, generate a new cookie. Once 2575 * generated, the socket cookie remains stable for the life of the 2576 * socket. This helper can be useful for monitoring per socket 2577 * networking traffic statistics as it provides a global socket 2578 * identifier that can be assumed unique. 2579 * Return 2580 * A 8-byte long unique number on success, or 0 if the socket 2581 * field is missing inside *skb*. 2582 * 2583 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2584 * Description 2585 * Equivalent to bpf_get_socket_cookie() helper that accepts 2586 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2587 * Return 2588 * A 8-byte long unique number. 2589 * 2590 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2591 * Description 2592 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2593 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2594 * Return 2595 * A 8-byte long unique number. 2596 * 2597 * u64 bpf_get_socket_cookie(struct sock *sk) 2598 * Description 2599 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2600 * *sk*, but gets socket from a BTF **struct sock**. This helper 2601 * also works for sleepable programs. 2602 * Return 2603 * A 8-byte long unique number or 0 if *sk* is NULL. 2604 * 2605 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2606 * Description 2607 * Get the owner UID of the socked associated to *skb*. 2608 * Return 2609 * The owner UID of the socket associated to *skb*. If the socket 2610 * is **NULL**, or if it is not a full socket (i.e. if it is a 2611 * time-wait or a request socket instead), **overflowuid** value 2612 * is returned (note that **overflowuid** might also be the actual 2613 * UID value for the socket). 2614 * 2615 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2616 * Description 2617 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2618 * to value *hash*. 2619 * Return 2620 * 0 2621 * 2622 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2623 * Description 2624 * Emulate a call to **setsockopt()** on the socket associated to 2625 * *bpf_socket*, which must be a full socket. The *level* at 2626 * which the option resides and the name *optname* of the option 2627 * must be specified, see **setsockopt(2)** for more information. 2628 * The option value of length *optlen* is pointed by *optval*. 2629 * 2630 * *bpf_socket* should be one of the following: 2631 * 2632 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2633 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2634 * and **BPF_CGROUP_INET6_CONNECT**. 2635 * 2636 * This helper actually implements a subset of **setsockopt()**. 2637 * It supports the following *level*\ s: 2638 * 2639 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2640 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2641 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2642 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2643 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2644 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2645 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2646 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2647 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2648 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2649 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2650 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2651 * **TCP_BPF_RTO_MIN**. 2652 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2653 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2654 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2655 * Return 2656 * 0 on success, or a negative error in case of failure. 2657 * 2658 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2659 * Description 2660 * Grow or shrink the room for data in the packet associated to 2661 * *skb* by *len_diff*, and according to the selected *mode*. 2662 * 2663 * By default, the helper will reset any offloaded checksum 2664 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2665 * by the following flag: 2666 * 2667 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2668 * checksum data of the skb to CHECKSUM_NONE. 2669 * 2670 * There are two supported modes at this time: 2671 * 2672 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2673 * (room space is added or removed between the layer 2 and 2674 * layer 3 headers). 2675 * 2676 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2677 * (room space is added or removed between the layer 3 and 2678 * layer 4 headers). 2679 * 2680 * The following flags are supported at this time: 2681 * 2682 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2683 * Adjusting mss in this way is not allowed for datagrams. 2684 * 2685 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2686 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2687 * Any new space is reserved to hold a tunnel header. 2688 * Configure skb offsets and other fields accordingly. 2689 * 2690 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2691 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2692 * Use with ENCAP_L3 flags to further specify the tunnel type. 2693 * 2694 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2695 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2696 * type; *len* is the length of the inner MAC header. 2697 * 2698 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2699 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2700 * L2 type as Ethernet. 2701 * 2702 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2703 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2704 * Indicate the new IP header version after decapsulating the outer 2705 * IP header. Used when the inner and outer IP versions are different. 2706 * 2707 * A call to this helper is susceptible to change the underlying 2708 * packet buffer. Therefore, at load time, all checks on pointers 2709 * previously done by the verifier are invalidated and must be 2710 * performed again, if the helper is used in combination with 2711 * direct packet access. 2712 * Return 2713 * 0 on success, or a negative error in case of failure. 2714 * 2715 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2716 * Description 2717 * Redirect the packet to the endpoint referenced by *map* at 2718 * index *key*. Depending on its type, this *map* can contain 2719 * references to net devices (for forwarding packets through other 2720 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2721 * but this is only implemented for native XDP (with driver 2722 * support) as of this writing). 2723 * 2724 * The lower two bits of *flags* are used as the return code if 2725 * the map lookup fails. This is so that the return value can be 2726 * one of the XDP program return codes up to **XDP_TX**, as chosen 2727 * by the caller. The higher bits of *flags* can be set to 2728 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2729 * 2730 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2731 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2732 * interface will be excluded when do broadcasting. 2733 * 2734 * See also **bpf_redirect**\ (), which only supports redirecting 2735 * to an ifindex, but doesn't require a map to do so. 2736 * Return 2737 * **XDP_REDIRECT** on success, or the value of the two lower bits 2738 * of the *flags* argument on error. 2739 * 2740 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2741 * Description 2742 * Redirect the packet to the socket referenced by *map* (of type 2743 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2744 * egress interfaces can be used for redirection. The 2745 * **BPF_F_INGRESS** value in *flags* is used to make the 2746 * distinction (ingress path is selected if the flag is present, 2747 * egress path otherwise). This is the only flag supported for now. 2748 * Return 2749 * **SK_PASS** on success, or **SK_DROP** on error. 2750 * 2751 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2752 * Description 2753 * Add an entry to, or update a *map* referencing sockets. The 2754 * *skops* is used as a new value for the entry associated to 2755 * *key*. *flags* is one of: 2756 * 2757 * **BPF_NOEXIST** 2758 * The entry for *key* must not exist in the map. 2759 * **BPF_EXIST** 2760 * The entry for *key* must already exist in the map. 2761 * **BPF_ANY** 2762 * No condition on the existence of the entry for *key*. 2763 * 2764 * If the *map* has eBPF programs (parser and verdict), those will 2765 * be inherited by the socket being added. If the socket is 2766 * already attached to eBPF programs, this results in an error. 2767 * Return 2768 * 0 on success, or a negative error in case of failure. 2769 * 2770 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2771 * Description 2772 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2773 * *delta* (which can be positive or negative). Note that this 2774 * operation modifies the address stored in *xdp_md*\ **->data**, 2775 * so the latter must be loaded only after the helper has been 2776 * called. 2777 * 2778 * The use of *xdp_md*\ **->data_meta** is optional and programs 2779 * are not required to use it. The rationale is that when the 2780 * packet is processed with XDP (e.g. as DoS filter), it is 2781 * possible to push further meta data along with it before passing 2782 * to the stack, and to give the guarantee that an ingress eBPF 2783 * program attached as a TC classifier on the same device can pick 2784 * this up for further post-processing. Since TC works with socket 2785 * buffers, it remains possible to set from XDP the **mark** or 2786 * **priority** pointers, or other pointers for the socket buffer. 2787 * Having this scratch space generic and programmable allows for 2788 * more flexibility as the user is free to store whatever meta 2789 * data they need. 2790 * 2791 * A call to this helper is susceptible to change the underlying 2792 * packet buffer. Therefore, at load time, all checks on pointers 2793 * previously done by the verifier are invalidated and must be 2794 * performed again, if the helper is used in combination with 2795 * direct packet access. 2796 * Return 2797 * 0 on success, or a negative error in case of failure. 2798 * 2799 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2800 * Description 2801 * Read the value of a perf event counter, and store it into *buf* 2802 * of size *buf_size*. This helper relies on a *map* of type 2803 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2804 * counter is selected when *map* is updated with perf event file 2805 * descriptors. The *map* is an array whose size is the number of 2806 * available CPUs, and each cell contains a value relative to one 2807 * CPU. The value to retrieve is indicated by *flags*, that 2808 * contains the index of the CPU to look up, masked with 2809 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2810 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2811 * current CPU should be retrieved. 2812 * 2813 * This helper behaves in a way close to 2814 * **bpf_perf_event_read**\ () helper, save that instead of 2815 * just returning the value observed, it fills the *buf* 2816 * structure. This allows for additional data to be retrieved: in 2817 * particular, the enabled and running times (in *buf*\ 2818 * **->enabled** and *buf*\ **->running**, respectively) are 2819 * copied. In general, **bpf_perf_event_read_value**\ () is 2820 * recommended over **bpf_perf_event_read**\ (), which has some 2821 * ABI issues and provides fewer functionalities. 2822 * 2823 * These values are interesting, because hardware PMU (Performance 2824 * Monitoring Unit) counters are limited resources. When there are 2825 * more PMU based perf events opened than available counters, 2826 * kernel will multiplex these events so each event gets certain 2827 * percentage (but not all) of the PMU time. In case that 2828 * multiplexing happens, the number of samples or counter value 2829 * will not reflect the case compared to when no multiplexing 2830 * occurs. This makes comparison between different runs difficult. 2831 * Typically, the counter value should be normalized before 2832 * comparing to other experiments. The usual normalization is done 2833 * as follows. 2834 * 2835 * :: 2836 * 2837 * normalized_counter = counter * t_enabled / t_running 2838 * 2839 * Where t_enabled is the time enabled for event and t_running is 2840 * the time running for event since last normalization. The 2841 * enabled and running times are accumulated since the perf event 2842 * open. To achieve scaling factor between two invocations of an 2843 * eBPF program, users can use CPU id as the key (which is 2844 * typical for perf array usage model) to remember the previous 2845 * value and do the calculation inside the eBPF program. 2846 * Return 2847 * 0 on success, or a negative error in case of failure. 2848 * 2849 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2850 * Description 2851 * For an eBPF program attached to a perf event, retrieve the 2852 * value of the event counter associated to *ctx* and store it in 2853 * the structure pointed by *buf* and of size *buf_size*. Enabled 2854 * and running times are also stored in the structure (see 2855 * description of helper **bpf_perf_event_read_value**\ () for 2856 * more details). 2857 * Return 2858 * 0 on success, or a negative error in case of failure. 2859 * 2860 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2861 * Description 2862 * Emulate a call to **getsockopt()** on the socket associated to 2863 * *bpf_socket*, which must be a full socket. The *level* at 2864 * which the option resides and the name *optname* of the option 2865 * must be specified, see **getsockopt(2)** for more information. 2866 * The retrieved value is stored in the structure pointed by 2867 * *opval* and of length *optlen*. 2868 * 2869 * *bpf_socket* should be one of the following: 2870 * 2871 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2872 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2873 * and **BPF_CGROUP_INET6_CONNECT**. 2874 * 2875 * This helper actually implements a subset of **getsockopt()**. 2876 * It supports the same set of *optname*\ s that is supported by 2877 * the **bpf_setsockopt**\ () helper. The exceptions are 2878 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 2879 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 2880 * Return 2881 * 0 on success, or a negative error in case of failure. 2882 * 2883 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2884 * Description 2885 * Used for error injection, this helper uses kprobes to override 2886 * the return value of the probed function, and to set it to *rc*. 2887 * The first argument is the context *regs* on which the kprobe 2888 * works. 2889 * 2890 * This helper works by setting the PC (program counter) 2891 * to an override function which is run in place of the original 2892 * probed function. This means the probed function is not run at 2893 * all. The replacement function just returns with the required 2894 * value. 2895 * 2896 * This helper has security implications, and thus is subject to 2897 * restrictions. It is only available if the kernel was compiled 2898 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2899 * option, and in this case it only works on functions tagged with 2900 * **ALLOW_ERROR_INJECTION** in the kernel code. 2901 * 2902 * Also, the helper is only available for the architectures having 2903 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2904 * x86 architecture is the only one to support this feature. 2905 * Return 2906 * 0 2907 * 2908 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2909 * Description 2910 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2911 * for the full TCP socket associated to *bpf_sock_ops* to 2912 * *argval*. 2913 * 2914 * The primary use of this field is to determine if there should 2915 * be calls to eBPF programs of type 2916 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2917 * code. A program of the same type can change its value, per 2918 * connection and as necessary, when the connection is 2919 * established. This field is directly accessible for reading, but 2920 * this helper must be used for updates in order to return an 2921 * error if an eBPF program tries to set a callback that is not 2922 * supported in the current kernel. 2923 * 2924 * *argval* is a flag array which can combine these flags: 2925 * 2926 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2927 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2928 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2929 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2930 * 2931 * Therefore, this function can be used to clear a callback flag by 2932 * setting the appropriate bit to zero. e.g. to disable the RTO 2933 * callback: 2934 * 2935 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2936 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2937 * 2938 * Here are some examples of where one could call such eBPF 2939 * program: 2940 * 2941 * * When RTO fires. 2942 * * When a packet is retransmitted. 2943 * * When the connection terminates. 2944 * * When a packet is sent. 2945 * * When a packet is received. 2946 * Return 2947 * Code **-EINVAL** if the socket is not a full TCP socket; 2948 * otherwise, a positive number containing the bits that could not 2949 * be set is returned (which comes down to 0 if all bits were set 2950 * as required). 2951 * 2952 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2953 * Description 2954 * This helper is used in programs implementing policies at the 2955 * socket level. If the message *msg* is allowed to pass (i.e. if 2956 * the verdict eBPF program returns **SK_PASS**), redirect it to 2957 * the socket referenced by *map* (of type 2958 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2959 * egress interfaces can be used for redirection. The 2960 * **BPF_F_INGRESS** value in *flags* is used to make the 2961 * distinction (ingress path is selected if the flag is present, 2962 * egress path otherwise). This is the only flag supported for now. 2963 * Return 2964 * **SK_PASS** on success, or **SK_DROP** on error. 2965 * 2966 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2967 * Description 2968 * For socket policies, apply the verdict of the eBPF program to 2969 * the next *bytes* (number of bytes) of message *msg*. 2970 * 2971 * For example, this helper can be used in the following cases: 2972 * 2973 * * A single **sendmsg**\ () or **sendfile**\ () system call 2974 * contains multiple logical messages that the eBPF program is 2975 * supposed to read and for which it should apply a verdict. 2976 * * An eBPF program only cares to read the first *bytes* of a 2977 * *msg*. If the message has a large payload, then setting up 2978 * and calling the eBPF program repeatedly for all bytes, even 2979 * though the verdict is already known, would create unnecessary 2980 * overhead. 2981 * 2982 * When called from within an eBPF program, the helper sets a 2983 * counter internal to the BPF infrastructure, that is used to 2984 * apply the last verdict to the next *bytes*. If *bytes* is 2985 * smaller than the current data being processed from a 2986 * **sendmsg**\ () or **sendfile**\ () system call, the first 2987 * *bytes* will be sent and the eBPF program will be re-run with 2988 * the pointer for start of data pointing to byte number *bytes* 2989 * **+ 1**. If *bytes* is larger than the current data being 2990 * processed, then the eBPF verdict will be applied to multiple 2991 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2992 * consumed. 2993 * 2994 * Note that if a socket closes with the internal counter holding 2995 * a non-zero value, this is not a problem because data is not 2996 * being buffered for *bytes* and is sent as it is received. 2997 * Return 2998 * 0 2999 * 3000 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3001 * Description 3002 * For socket policies, prevent the execution of the verdict eBPF 3003 * program for message *msg* until *bytes* (byte number) have been 3004 * accumulated. 3005 * 3006 * This can be used when one needs a specific number of bytes 3007 * before a verdict can be assigned, even if the data spans 3008 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3009 * case would be a user calling **sendmsg**\ () repeatedly with 3010 * 1-byte long message segments. Obviously, this is bad for 3011 * performance, but it is still valid. If the eBPF program needs 3012 * *bytes* bytes to validate a header, this helper can be used to 3013 * prevent the eBPF program to be called again until *bytes* have 3014 * been accumulated. 3015 * Return 3016 * 0 3017 * 3018 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3019 * Description 3020 * For socket policies, pull in non-linear data from user space 3021 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3022 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3023 * respectively. 3024 * 3025 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3026 * *msg* it can only parse data that the (**data**, **data_end**) 3027 * pointers have already consumed. For **sendmsg**\ () hooks this 3028 * is likely the first scatterlist element. But for calls relying 3029 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3030 * be the range (**0**, **0**) because the data is shared with 3031 * user space and by default the objective is to avoid allowing 3032 * user space to modify data while (or after) eBPF verdict is 3033 * being decided. This helper can be used to pull in data and to 3034 * set the start and end pointer to given values. Data will be 3035 * copied if necessary (i.e. if data was not linear and if start 3036 * and end pointers do not point to the same chunk). 3037 * 3038 * A call to this helper is susceptible to change the underlying 3039 * packet buffer. Therefore, at load time, all checks on pointers 3040 * previously done by the verifier are invalidated and must be 3041 * performed again, if the helper is used in combination with 3042 * direct packet access. 3043 * 3044 * All values for *flags* are reserved for future usage, and must 3045 * be left at zero. 3046 * Return 3047 * 0 on success, or a negative error in case of failure. 3048 * 3049 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3050 * Description 3051 * Bind the socket associated to *ctx* to the address pointed by 3052 * *addr*, of length *addr_len*. This allows for making outgoing 3053 * connection from the desired IP address, which can be useful for 3054 * example when all processes inside a cgroup should use one 3055 * single IP address on a host that has multiple IP configured. 3056 * 3057 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3058 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3059 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3060 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3061 * behavior and lets the kernel efficiently pick up an unused 3062 * port as long as 4-tuple is unique. Passing non-zero port might 3063 * lead to degraded performance. 3064 * Return 3065 * 0 on success, or a negative error in case of failure. 3066 * 3067 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3068 * Description 3069 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3070 * possible to both shrink and grow the packet tail. 3071 * Shrink done via *delta* being a negative integer. 3072 * 3073 * A call to this helper is susceptible to change the underlying 3074 * packet buffer. Therefore, at load time, all checks on pointers 3075 * previously done by the verifier are invalidated and must be 3076 * performed again, if the helper is used in combination with 3077 * direct packet access. 3078 * Return 3079 * 0 on success, or a negative error in case of failure. 3080 * 3081 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3082 * Description 3083 * Retrieve the XFRM state (IP transform framework, see also 3084 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3085 * 3086 * The retrieved value is stored in the **struct bpf_xfrm_state** 3087 * pointed by *xfrm_state* and of length *size*. 3088 * 3089 * All values for *flags* are reserved for future usage, and must 3090 * be left at zero. 3091 * 3092 * This helper is available only if the kernel was compiled with 3093 * **CONFIG_XFRM** configuration option. 3094 * Return 3095 * 0 on success, or a negative error in case of failure. 3096 * 3097 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3098 * Description 3099 * Return a user or a kernel stack in bpf program provided buffer. 3100 * To achieve this, the helper needs *ctx*, which is a pointer 3101 * to the context on which the tracing program is executed. 3102 * To store the stacktrace, the bpf program provides *buf* with 3103 * a nonnegative *size*. 3104 * 3105 * The last argument, *flags*, holds the number of stack frames to 3106 * skip (from 0 to 255), masked with 3107 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3108 * the following flags: 3109 * 3110 * **BPF_F_USER_STACK** 3111 * Collect a user space stack instead of a kernel stack. 3112 * **BPF_F_USER_BUILD_ID** 3113 * Collect (build_id, file_offset) instead of ips for user 3114 * stack, only valid if **BPF_F_USER_STACK** is also 3115 * specified. 3116 * 3117 * *file_offset* is an offset relative to the beginning 3118 * of the executable or shared object file backing the vma 3119 * which the *ip* falls in. It is *not* an offset relative 3120 * to that object's base address. Accordingly, it must be 3121 * adjusted by adding (sh_addr - sh_offset), where 3122 * sh_{addr,offset} correspond to the executable section 3123 * containing *file_offset* in the object, for comparisons 3124 * to symbols' st_value to be valid. 3125 * 3126 * **bpf_get_stack**\ () can collect up to 3127 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3128 * to sufficient large buffer size. Note that 3129 * this limit can be controlled with the **sysctl** program, and 3130 * that it should be manually increased in order to profile long 3131 * user stacks (such as stacks for Java programs). To do so, use: 3132 * 3133 * :: 3134 * 3135 * # sysctl kernel.perf_event_max_stack=<new value> 3136 * Return 3137 * The non-negative copied *buf* length equal to or less than 3138 * *size* on success, or a negative error in case of failure. 3139 * 3140 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3141 * Description 3142 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3143 * it provides an easy way to load *len* bytes from *offset* 3144 * from the packet associated to *skb*, into the buffer pointed 3145 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3146 * a fifth argument *start_header* exists in order to select a 3147 * base offset to start from. *start_header* can be one of: 3148 * 3149 * **BPF_HDR_START_MAC** 3150 * Base offset to load data from is *skb*'s mac header. 3151 * **BPF_HDR_START_NET** 3152 * Base offset to load data from is *skb*'s network header. 3153 * 3154 * In general, "direct packet access" is the preferred method to 3155 * access packet data, however, this helper is in particular useful 3156 * in socket filters where *skb*\ **->data** does not always point 3157 * to the start of the mac header and where "direct packet access" 3158 * is not available. 3159 * Return 3160 * 0 on success, or a negative error in case of failure. 3161 * 3162 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3163 * Description 3164 * Do FIB lookup in kernel tables using parameters in *params*. 3165 * If lookup is successful and result shows packet is to be 3166 * forwarded, the neighbor tables are searched for the nexthop. 3167 * If successful (ie., FIB lookup shows forwarding and nexthop 3168 * is resolved), the nexthop address is returned in ipv4_dst 3169 * or ipv6_dst based on family, smac is set to mac address of 3170 * egress device, dmac is set to nexthop mac address, rt_metric 3171 * is set to metric from route (IPv4/IPv6 only), and ifindex 3172 * is set to the device index of the nexthop from the FIB lookup. 3173 * 3174 * *plen* argument is the size of the passed in struct. 3175 * *flags* argument can be a combination of one or more of the 3176 * following values: 3177 * 3178 * **BPF_FIB_LOOKUP_DIRECT** 3179 * Do a direct table lookup vs full lookup using FIB 3180 * rules. 3181 * **BPF_FIB_LOOKUP_TBID** 3182 * Used with BPF_FIB_LOOKUP_DIRECT. 3183 * Use the routing table ID present in *params*->tbid 3184 * for the fib lookup. 3185 * **BPF_FIB_LOOKUP_OUTPUT** 3186 * Perform lookup from an egress perspective (default is 3187 * ingress). 3188 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3189 * Skip the neighbour table lookup. *params*->dmac 3190 * and *params*->smac will not be set as output. A common 3191 * use case is to call **bpf_redirect_neigh**\ () after 3192 * doing **bpf_fib_lookup**\ (). 3193 * 3194 * *ctx* is either **struct xdp_md** for XDP programs or 3195 * **struct sk_buff** tc cls_act programs. 3196 * Return 3197 * * < 0 if any input argument is invalid 3198 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3199 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3200 * packet is not forwarded or needs assist from full stack 3201 * 3202 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3203 * was exceeded and output params->mtu_result contains the MTU. 3204 * 3205 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3206 * Description 3207 * Add an entry to, or update a sockhash *map* referencing sockets. 3208 * The *skops* is used as a new value for the entry associated to 3209 * *key*. *flags* is one of: 3210 * 3211 * **BPF_NOEXIST** 3212 * The entry for *key* must not exist in the map. 3213 * **BPF_EXIST** 3214 * The entry for *key* must already exist in the map. 3215 * **BPF_ANY** 3216 * No condition on the existence of the entry for *key*. 3217 * 3218 * If the *map* has eBPF programs (parser and verdict), those will 3219 * be inherited by the socket being added. If the socket is 3220 * already attached to eBPF programs, this results in an error. 3221 * Return 3222 * 0 on success, or a negative error in case of failure. 3223 * 3224 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3225 * Description 3226 * This helper is used in programs implementing policies at the 3227 * socket level. If the message *msg* is allowed to pass (i.e. if 3228 * the verdict eBPF program returns **SK_PASS**), redirect it to 3229 * the socket referenced by *map* (of type 3230 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3231 * egress interfaces can be used for redirection. The 3232 * **BPF_F_INGRESS** value in *flags* is used to make the 3233 * distinction (ingress path is selected if the flag is present, 3234 * egress path otherwise). This is the only flag supported for now. 3235 * Return 3236 * **SK_PASS** on success, or **SK_DROP** on error. 3237 * 3238 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3239 * Description 3240 * This helper is used in programs implementing policies at the 3241 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3242 * if the verdict eBPF program returns **SK_PASS**), redirect it 3243 * to the socket referenced by *map* (of type 3244 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3245 * egress interfaces can be used for redirection. The 3246 * **BPF_F_INGRESS** value in *flags* is used to make the 3247 * distinction (ingress path is selected if the flag is present, 3248 * egress otherwise). This is the only flag supported for now. 3249 * Return 3250 * **SK_PASS** on success, or **SK_DROP** on error. 3251 * 3252 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3253 * Description 3254 * Encapsulate the packet associated to *skb* within a Layer 3 3255 * protocol header. This header is provided in the buffer at 3256 * address *hdr*, with *len* its size in bytes. *type* indicates 3257 * the protocol of the header and can be one of: 3258 * 3259 * **BPF_LWT_ENCAP_SEG6** 3260 * IPv6 encapsulation with Segment Routing Header 3261 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3262 * the IPv6 header is computed by the kernel. 3263 * **BPF_LWT_ENCAP_SEG6_INLINE** 3264 * Only works if *skb* contains an IPv6 packet. Insert a 3265 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3266 * the IPv6 header. 3267 * **BPF_LWT_ENCAP_IP** 3268 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3269 * must be IPv4 or IPv6, followed by zero or more 3270 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3271 * total bytes in all prepended headers. Please note that 3272 * if **skb_is_gso**\ (*skb*) is true, no more than two 3273 * headers can be prepended, and the inner header, if 3274 * present, should be either GRE or UDP/GUE. 3275 * 3276 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3277 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3278 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3279 * **BPF_PROG_TYPE_LWT_XMIT**. 3280 * 3281 * A call to this helper is susceptible to change the underlying 3282 * packet buffer. Therefore, at load time, all checks on pointers 3283 * previously done by the verifier are invalidated and must be 3284 * performed again, if the helper is used in combination with 3285 * direct packet access. 3286 * Return 3287 * 0 on success, or a negative error in case of failure. 3288 * 3289 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3290 * Description 3291 * Store *len* bytes from address *from* into the packet 3292 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3293 * inside the outermost IPv6 Segment Routing Header can be 3294 * modified through this helper. 3295 * 3296 * A call to this helper is susceptible to change the underlying 3297 * packet buffer. Therefore, at load time, all checks on pointers 3298 * previously done by the verifier are invalidated and must be 3299 * performed again, if the helper is used in combination with 3300 * direct packet access. 3301 * Return 3302 * 0 on success, or a negative error in case of failure. 3303 * 3304 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3305 * Description 3306 * Adjust the size allocated to TLVs in the outermost IPv6 3307 * Segment Routing Header contained in the packet associated to 3308 * *skb*, at position *offset* by *delta* bytes. Only offsets 3309 * after the segments are accepted. *delta* can be as well 3310 * positive (growing) as negative (shrinking). 3311 * 3312 * A call to this helper is susceptible to change the underlying 3313 * packet buffer. Therefore, at load time, all checks on pointers 3314 * previously done by the verifier are invalidated and must be 3315 * performed again, if the helper is used in combination with 3316 * direct packet access. 3317 * Return 3318 * 0 on success, or a negative error in case of failure. 3319 * 3320 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3321 * Description 3322 * Apply an IPv6 Segment Routing action of type *action* to the 3323 * packet associated to *skb*. Each action takes a parameter 3324 * contained at address *param*, and of length *param_len* bytes. 3325 * *action* can be one of: 3326 * 3327 * **SEG6_LOCAL_ACTION_END_X** 3328 * End.X action: Endpoint with Layer-3 cross-connect. 3329 * Type of *param*: **struct in6_addr**. 3330 * **SEG6_LOCAL_ACTION_END_T** 3331 * End.T action: Endpoint with specific IPv6 table lookup. 3332 * Type of *param*: **int**. 3333 * **SEG6_LOCAL_ACTION_END_B6** 3334 * End.B6 action: Endpoint bound to an SRv6 policy. 3335 * Type of *param*: **struct ipv6_sr_hdr**. 3336 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3337 * End.B6.Encap action: Endpoint bound to an SRv6 3338 * encapsulation policy. 3339 * Type of *param*: **struct ipv6_sr_hdr**. 3340 * 3341 * A call to this helper is susceptible to change the underlying 3342 * packet buffer. Therefore, at load time, all checks on pointers 3343 * previously done by the verifier are invalidated and must be 3344 * performed again, if the helper is used in combination with 3345 * direct packet access. 3346 * Return 3347 * 0 on success, or a negative error in case of failure. 3348 * 3349 * long bpf_rc_repeat(void *ctx) 3350 * Description 3351 * This helper is used in programs implementing IR decoding, to 3352 * report a successfully decoded repeat key message. This delays 3353 * the generation of a key up event for previously generated 3354 * key down event. 3355 * 3356 * Some IR protocols like NEC have a special IR message for 3357 * repeating last button, for when a button is held down. 3358 * 3359 * The *ctx* should point to the lirc sample as passed into 3360 * the program. 3361 * 3362 * This helper is only available is the kernel was compiled with 3363 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3364 * "**y**". 3365 * Return 3366 * 0 3367 * 3368 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3369 * Description 3370 * This helper is used in programs implementing IR decoding, to 3371 * report a successfully decoded key press with *scancode*, 3372 * *toggle* value in the given *protocol*. The scancode will be 3373 * translated to a keycode using the rc keymap, and reported as 3374 * an input key down event. After a period a key up event is 3375 * generated. This period can be extended by calling either 3376 * **bpf_rc_keydown**\ () again with the same values, or calling 3377 * **bpf_rc_repeat**\ (). 3378 * 3379 * Some protocols include a toggle bit, in case the button was 3380 * released and pressed again between consecutive scancodes. 3381 * 3382 * The *ctx* should point to the lirc sample as passed into 3383 * the program. 3384 * 3385 * The *protocol* is the decoded protocol number (see 3386 * **enum rc_proto** for some predefined values). 3387 * 3388 * This helper is only available is the kernel was compiled with 3389 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3390 * "**y**". 3391 * Return 3392 * 0 3393 * 3394 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3395 * Description 3396 * Return the cgroup v2 id of the socket associated with the *skb*. 3397 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3398 * helper for cgroup v1 by providing a tag resp. identifier that 3399 * can be matched on or used for map lookups e.g. to implement 3400 * policy. The cgroup v2 id of a given path in the hierarchy is 3401 * exposed in user space through the f_handle API in order to get 3402 * to the same 64-bit id. 3403 * 3404 * This helper can be used on TC egress path, but not on ingress, 3405 * and is available only if the kernel was compiled with the 3406 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3407 * Return 3408 * The id is returned or 0 in case the id could not be retrieved. 3409 * 3410 * u64 bpf_get_current_cgroup_id(void) 3411 * Description 3412 * Get the current cgroup id based on the cgroup within which 3413 * the current task is running. 3414 * Return 3415 * A 64-bit integer containing the current cgroup id based 3416 * on the cgroup within which the current task is running. 3417 * 3418 * void *bpf_get_local_storage(void *map, u64 flags) 3419 * Description 3420 * Get the pointer to the local storage area. 3421 * The type and the size of the local storage is defined 3422 * by the *map* argument. 3423 * The *flags* meaning is specific for each map type, 3424 * and has to be 0 for cgroup local storage. 3425 * 3426 * Depending on the BPF program type, a local storage area 3427 * can be shared between multiple instances of the BPF program, 3428 * running simultaneously. 3429 * 3430 * A user should care about the synchronization by himself. 3431 * For example, by using the **BPF_ATOMIC** instructions to alter 3432 * the shared data. 3433 * Return 3434 * A pointer to the local storage area. 3435 * 3436 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3437 * Description 3438 * Select a **SO_REUSEPORT** socket from a 3439 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3440 * It checks the selected socket is matching the incoming 3441 * request in the socket buffer. 3442 * Return 3443 * 0 on success, or a negative error in case of failure. 3444 * 3445 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3446 * Description 3447 * Return id of cgroup v2 that is ancestor of cgroup associated 3448 * with the *skb* at the *ancestor_level*. The root cgroup is at 3449 * *ancestor_level* zero and each step down the hierarchy 3450 * increments the level. If *ancestor_level* == level of cgroup 3451 * associated with *skb*, then return value will be same as that 3452 * of **bpf_skb_cgroup_id**\ (). 3453 * 3454 * The helper is useful to implement policies based on cgroups 3455 * that are upper in hierarchy than immediate cgroup associated 3456 * with *skb*. 3457 * 3458 * The format of returned id and helper limitations are same as in 3459 * **bpf_skb_cgroup_id**\ (). 3460 * Return 3461 * The id is returned or 0 in case the id could not be retrieved. 3462 * 3463 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3464 * Description 3465 * Look for TCP socket matching *tuple*, optionally in a child 3466 * network namespace *netns*. The return value must be checked, 3467 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3468 * 3469 * The *ctx* should point to the context of the program, such as 3470 * the skb or socket (depending on the hook in use). This is used 3471 * to determine the base network namespace for the lookup. 3472 * 3473 * *tuple_size* must be one of: 3474 * 3475 * **sizeof**\ (*tuple*\ **->ipv4**) 3476 * Look for an IPv4 socket. 3477 * **sizeof**\ (*tuple*\ **->ipv6**) 3478 * Look for an IPv6 socket. 3479 * 3480 * If the *netns* is a negative signed 32-bit integer, then the 3481 * socket lookup table in the netns associated with the *ctx* 3482 * will be used. For the TC hooks, this is the netns of the device 3483 * in the skb. For socket hooks, this is the netns of the socket. 3484 * If *netns* is any other signed 32-bit value greater than or 3485 * equal to zero then it specifies the ID of the netns relative to 3486 * the netns associated with the *ctx*. *netns* values beyond the 3487 * range of 32-bit integers are reserved for future use. 3488 * 3489 * All values for *flags* are reserved for future usage, and must 3490 * be left at zero. 3491 * 3492 * This helper is available only if the kernel was compiled with 3493 * **CONFIG_NET** configuration option. 3494 * Return 3495 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3496 * For sockets with reuseport option, the **struct bpf_sock** 3497 * result is from *reuse*\ **->socks**\ [] using the hash of the 3498 * tuple. 3499 * 3500 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3501 * Description 3502 * Look for UDP socket matching *tuple*, optionally in a child 3503 * network namespace *netns*. The return value must be checked, 3504 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3505 * 3506 * The *ctx* should point to the context of the program, such as 3507 * the skb or socket (depending on the hook in use). This is used 3508 * to determine the base network namespace for the lookup. 3509 * 3510 * *tuple_size* must be one of: 3511 * 3512 * **sizeof**\ (*tuple*\ **->ipv4**) 3513 * Look for an IPv4 socket. 3514 * **sizeof**\ (*tuple*\ **->ipv6**) 3515 * Look for an IPv6 socket. 3516 * 3517 * If the *netns* is a negative signed 32-bit integer, then the 3518 * socket lookup table in the netns associated with the *ctx* 3519 * will be used. For the TC hooks, this is the netns of the device 3520 * in the skb. For socket hooks, this is the netns of the socket. 3521 * If *netns* is any other signed 32-bit value greater than or 3522 * equal to zero then it specifies the ID of the netns relative to 3523 * the netns associated with the *ctx*. *netns* values beyond the 3524 * range of 32-bit integers are reserved for future use. 3525 * 3526 * All values for *flags* are reserved for future usage, and must 3527 * be left at zero. 3528 * 3529 * This helper is available only if the kernel was compiled with 3530 * **CONFIG_NET** configuration option. 3531 * Return 3532 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3533 * For sockets with reuseport option, the **struct bpf_sock** 3534 * result is from *reuse*\ **->socks**\ [] using the hash of the 3535 * tuple. 3536 * 3537 * long bpf_sk_release(void *sock) 3538 * Description 3539 * Release the reference held by *sock*. *sock* must be a 3540 * non-**NULL** pointer that was returned from 3541 * **bpf_sk_lookup_xxx**\ (). 3542 * Return 3543 * 0 on success, or a negative error in case of failure. 3544 * 3545 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3546 * Description 3547 * Push an element *value* in *map*. *flags* is one of: 3548 * 3549 * **BPF_EXIST** 3550 * If the queue/stack is full, the oldest element is 3551 * removed to make room for this. 3552 * Return 3553 * 0 on success, or a negative error in case of failure. 3554 * 3555 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3556 * Description 3557 * Pop an element from *map*. 3558 * Return 3559 * 0 on success, or a negative error in case of failure. 3560 * 3561 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3562 * Description 3563 * Get an element from *map* without removing it. 3564 * Return 3565 * 0 on success, or a negative error in case of failure. 3566 * 3567 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3568 * Description 3569 * For socket policies, insert *len* bytes into *msg* at offset 3570 * *start*. 3571 * 3572 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3573 * *msg* it may want to insert metadata or options into the *msg*. 3574 * This can later be read and used by any of the lower layer BPF 3575 * hooks. 3576 * 3577 * This helper may fail if under memory pressure (a malloc 3578 * fails) in these cases BPF programs will get an appropriate 3579 * error and BPF programs will need to handle them. 3580 * Return 3581 * 0 on success, or a negative error in case of failure. 3582 * 3583 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3584 * Description 3585 * Will remove *len* bytes from a *msg* starting at byte *start*. 3586 * This may result in **ENOMEM** errors under certain situations if 3587 * an allocation and copy are required due to a full ring buffer. 3588 * However, the helper will try to avoid doing the allocation 3589 * if possible. Other errors can occur if input parameters are 3590 * invalid either due to *start* byte not being valid part of *msg* 3591 * payload and/or *pop* value being to large. 3592 * Return 3593 * 0 on success, or a negative error in case of failure. 3594 * 3595 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3596 * Description 3597 * This helper is used in programs implementing IR decoding, to 3598 * report a successfully decoded pointer movement. 3599 * 3600 * The *ctx* should point to the lirc sample as passed into 3601 * the program. 3602 * 3603 * This helper is only available is the kernel was compiled with 3604 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3605 * "**y**". 3606 * Return 3607 * 0 3608 * 3609 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3610 * Description 3611 * Acquire a spinlock represented by the pointer *lock*, which is 3612 * stored as part of a value of a map. Taking the lock allows to 3613 * safely update the rest of the fields in that value. The 3614 * spinlock can (and must) later be released with a call to 3615 * **bpf_spin_unlock**\ (\ *lock*\ ). 3616 * 3617 * Spinlocks in BPF programs come with a number of restrictions 3618 * and constraints: 3619 * 3620 * * **bpf_spin_lock** objects are only allowed inside maps of 3621 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3622 * list could be extended in the future). 3623 * * BTF description of the map is mandatory. 3624 * * The BPF program can take ONE lock at a time, since taking two 3625 * or more could cause dead locks. 3626 * * Only one **struct bpf_spin_lock** is allowed per map element. 3627 * * When the lock is taken, calls (either BPF to BPF or helpers) 3628 * are not allowed. 3629 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3630 * allowed inside a spinlock-ed region. 3631 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3632 * the lock, on all execution paths, before it returns. 3633 * * The BPF program can access **struct bpf_spin_lock** only via 3634 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3635 * helpers. Loading or storing data into the **struct 3636 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3637 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3638 * of the map value must be a struct and have **struct 3639 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3640 * Nested lock inside another struct is not allowed. 3641 * * The **struct bpf_spin_lock** *lock* field in a map value must 3642 * be aligned on a multiple of 4 bytes in that value. 3643 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3644 * the **bpf_spin_lock** field to user space. 3645 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3646 * a BPF program, do not update the **bpf_spin_lock** field. 3647 * * **bpf_spin_lock** cannot be on the stack or inside a 3648 * networking packet (it can only be inside of a map values). 3649 * * **bpf_spin_lock** is available to root only. 3650 * * Tracing programs and socket filter programs cannot use 3651 * **bpf_spin_lock**\ () due to insufficient preemption checks 3652 * (but this may change in the future). 3653 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3654 * Return 3655 * 0 3656 * 3657 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3658 * Description 3659 * Release the *lock* previously locked by a call to 3660 * **bpf_spin_lock**\ (\ *lock*\ ). 3661 * Return 3662 * 0 3663 * 3664 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3665 * Description 3666 * This helper gets a **struct bpf_sock** pointer such 3667 * that all the fields in this **bpf_sock** can be accessed. 3668 * Return 3669 * A **struct bpf_sock** pointer on success, or **NULL** in 3670 * case of failure. 3671 * 3672 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3673 * Description 3674 * This helper gets a **struct bpf_tcp_sock** pointer from a 3675 * **struct bpf_sock** pointer. 3676 * Return 3677 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3678 * case of failure. 3679 * 3680 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3681 * Description 3682 * Set ECN (Explicit Congestion Notification) field of IP header 3683 * to **CE** (Congestion Encountered) if current value is **ECT** 3684 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3685 * and IPv4. 3686 * Return 3687 * 1 if the **CE** flag is set (either by the current helper call 3688 * or because it was already present), 0 if it is not set. 3689 * 3690 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3691 * Description 3692 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3693 * **bpf_sk_release**\ () is unnecessary and not allowed. 3694 * Return 3695 * A **struct bpf_sock** pointer on success, or **NULL** in 3696 * case of failure. 3697 * 3698 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3699 * Description 3700 * Look for TCP socket matching *tuple*, optionally in a child 3701 * network namespace *netns*. The return value must be checked, 3702 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3703 * 3704 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3705 * that it also returns timewait or request sockets. Use 3706 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3707 * full structure. 3708 * 3709 * This helper is available only if the kernel was compiled with 3710 * **CONFIG_NET** configuration option. 3711 * Return 3712 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3713 * For sockets with reuseport option, the **struct bpf_sock** 3714 * result is from *reuse*\ **->socks**\ [] using the hash of the 3715 * tuple. 3716 * 3717 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3718 * Description 3719 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3720 * the listening socket in *sk*. 3721 * 3722 * *iph* points to the start of the IPv4 or IPv6 header, while 3723 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3724 * **sizeof**\ (**struct ipv6hdr**). 3725 * 3726 * *th* points to the start of the TCP header, while *th_len* 3727 * contains the length of the TCP header (at least 3728 * **sizeof**\ (**struct tcphdr**)). 3729 * Return 3730 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3731 * error otherwise. 3732 * 3733 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3734 * Description 3735 * Get name of sysctl in /proc/sys/ and copy it into provided by 3736 * program buffer *buf* of size *buf_len*. 3737 * 3738 * The buffer is always NUL terminated, unless it's zero-sized. 3739 * 3740 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3741 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3742 * only (e.g. "tcp_mem"). 3743 * Return 3744 * Number of character copied (not including the trailing NUL). 3745 * 3746 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3747 * truncated name in this case). 3748 * 3749 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3750 * Description 3751 * Get current value of sysctl as it is presented in /proc/sys 3752 * (incl. newline, etc), and copy it as a string into provided 3753 * by program buffer *buf* of size *buf_len*. 3754 * 3755 * The whole value is copied, no matter what file position user 3756 * space issued e.g. sys_read at. 3757 * 3758 * The buffer is always NUL terminated, unless it's zero-sized. 3759 * Return 3760 * Number of character copied (not including the trailing NUL). 3761 * 3762 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3763 * truncated name in this case). 3764 * 3765 * **-EINVAL** if current value was unavailable, e.g. because 3766 * sysctl is uninitialized and read returns -EIO for it. 3767 * 3768 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3769 * Description 3770 * Get new value being written by user space to sysctl (before 3771 * the actual write happens) and copy it as a string into 3772 * provided by program buffer *buf* of size *buf_len*. 3773 * 3774 * User space may write new value at file position > 0. 3775 * 3776 * The buffer is always NUL terminated, unless it's zero-sized. 3777 * Return 3778 * Number of character copied (not including the trailing NUL). 3779 * 3780 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3781 * truncated name in this case). 3782 * 3783 * **-EINVAL** if sysctl is being read. 3784 * 3785 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3786 * Description 3787 * Override new value being written by user space to sysctl with 3788 * value provided by program in buffer *buf* of size *buf_len*. 3789 * 3790 * *buf* should contain a string in same form as provided by user 3791 * space on sysctl write. 3792 * 3793 * User space may write new value at file position > 0. To override 3794 * the whole sysctl value file position should be set to zero. 3795 * Return 3796 * 0 on success. 3797 * 3798 * **-E2BIG** if the *buf_len* is too big. 3799 * 3800 * **-EINVAL** if sysctl is being read. 3801 * 3802 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3803 * Description 3804 * Convert the initial part of the string from buffer *buf* of 3805 * size *buf_len* to a long integer according to the given base 3806 * and save the result in *res*. 3807 * 3808 * The string may begin with an arbitrary amount of white space 3809 * (as determined by **isspace**\ (3)) followed by a single 3810 * optional '**-**' sign. 3811 * 3812 * Five least significant bits of *flags* encode base, other bits 3813 * are currently unused. 3814 * 3815 * Base must be either 8, 10, 16 or 0 to detect it automatically 3816 * similar to user space **strtol**\ (3). 3817 * Return 3818 * Number of characters consumed on success. Must be positive but 3819 * no more than *buf_len*. 3820 * 3821 * **-EINVAL** if no valid digits were found or unsupported base 3822 * was provided. 3823 * 3824 * **-ERANGE** if resulting value was out of range. 3825 * 3826 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3827 * Description 3828 * Convert the initial part of the string from buffer *buf* of 3829 * size *buf_len* to an unsigned long integer according to the 3830 * given base and save the result in *res*. 3831 * 3832 * The string may begin with an arbitrary amount of white space 3833 * (as determined by **isspace**\ (3)). 3834 * 3835 * Five least significant bits of *flags* encode base, other bits 3836 * are currently unused. 3837 * 3838 * Base must be either 8, 10, 16 or 0 to detect it automatically 3839 * similar to user space **strtoul**\ (3). 3840 * Return 3841 * Number of characters consumed on success. Must be positive but 3842 * no more than *buf_len*. 3843 * 3844 * **-EINVAL** if no valid digits were found or unsupported base 3845 * was provided. 3846 * 3847 * **-ERANGE** if resulting value was out of range. 3848 * 3849 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3850 * Description 3851 * Get a bpf-local-storage from a *sk*. 3852 * 3853 * Logically, it could be thought of getting the value from 3854 * a *map* with *sk* as the **key**. From this 3855 * perspective, the usage is not much different from 3856 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3857 * helper enforces the key must be a full socket and the map must 3858 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3859 * 3860 * Underneath, the value is stored locally at *sk* instead of 3861 * the *map*. The *map* is used as the bpf-local-storage 3862 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3863 * searched against all bpf-local-storages residing at *sk*. 3864 * 3865 * *sk* is a kernel **struct sock** pointer for LSM program. 3866 * *sk* is a **struct bpf_sock** pointer for other program types. 3867 * 3868 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3869 * used such that a new bpf-local-storage will be 3870 * created if one does not exist. *value* can be used 3871 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3872 * the initial value of a bpf-local-storage. If *value* is 3873 * **NULL**, the new bpf-local-storage will be zero initialized. 3874 * Return 3875 * A bpf-local-storage pointer is returned on success. 3876 * 3877 * **NULL** if not found or there was an error in adding 3878 * a new bpf-local-storage. 3879 * 3880 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3881 * Description 3882 * Delete a bpf-local-storage from a *sk*. 3883 * Return 3884 * 0 on success. 3885 * 3886 * **-ENOENT** if the bpf-local-storage cannot be found. 3887 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3888 * 3889 * long bpf_send_signal(u32 sig) 3890 * Description 3891 * Send signal *sig* to the process of the current task. 3892 * The signal may be delivered to any of this process's threads. 3893 * Return 3894 * 0 on success or successfully queued. 3895 * 3896 * **-EBUSY** if work queue under nmi is full. 3897 * 3898 * **-EINVAL** if *sig* is invalid. 3899 * 3900 * **-EPERM** if no permission to send the *sig*. 3901 * 3902 * **-EAGAIN** if bpf program can try again. 3903 * 3904 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3905 * Description 3906 * Try to issue a SYN cookie for the packet with corresponding 3907 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3908 * 3909 * *iph* points to the start of the IPv4 or IPv6 header, while 3910 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3911 * **sizeof**\ (**struct ipv6hdr**). 3912 * 3913 * *th* points to the start of the TCP header, while *th_len* 3914 * contains the length of the TCP header with options (at least 3915 * **sizeof**\ (**struct tcphdr**)). 3916 * Return 3917 * On success, lower 32 bits hold the generated SYN cookie in 3918 * followed by 16 bits which hold the MSS value for that cookie, 3919 * and the top 16 bits are unused. 3920 * 3921 * On failure, the returned value is one of the following: 3922 * 3923 * **-EINVAL** SYN cookie cannot be issued due to error 3924 * 3925 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3926 * 3927 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3928 * 3929 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3930 * 3931 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3932 * Description 3933 * Write raw *data* blob into a special BPF perf event held by 3934 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3935 * event must have the following attributes: **PERF_SAMPLE_RAW** 3936 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3937 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3938 * 3939 * The *flags* are used to indicate the index in *map* for which 3940 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3941 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3942 * to indicate that the index of the current CPU core should be 3943 * used. 3944 * 3945 * The value to write, of *size*, is passed through eBPF stack and 3946 * pointed by *data*. 3947 * 3948 * *ctx* is a pointer to in-kernel struct sk_buff. 3949 * 3950 * This helper is similar to **bpf_perf_event_output**\ () but 3951 * restricted to raw_tracepoint bpf programs. 3952 * Return 3953 * 0 on success, or a negative error in case of failure. 3954 * 3955 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3956 * Description 3957 * Safely attempt to read *size* bytes from user space address 3958 * *unsafe_ptr* and store the data in *dst*. 3959 * Return 3960 * 0 on success, or a negative error in case of failure. 3961 * 3962 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3963 * Description 3964 * Safely attempt to read *size* bytes from kernel space address 3965 * *unsafe_ptr* and store the data in *dst*. 3966 * Return 3967 * 0 on success, or a negative error in case of failure. 3968 * 3969 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3970 * Description 3971 * Copy a NUL terminated string from an unsafe user address 3972 * *unsafe_ptr* to *dst*. The *size* should include the 3973 * terminating NUL byte. In case the string length is smaller than 3974 * *size*, the target is not padded with further NUL bytes. If the 3975 * string length is larger than *size*, just *size*-1 bytes are 3976 * copied and the last byte is set to NUL. 3977 * 3978 * On success, returns the number of bytes that were written, 3979 * including the terminal NUL. This makes this helper useful in 3980 * tracing programs for reading strings, and more importantly to 3981 * get its length at runtime. See the following snippet: 3982 * 3983 * :: 3984 * 3985 * SEC("kprobe/sys_open") 3986 * void bpf_sys_open(struct pt_regs *ctx) 3987 * { 3988 * char buf[PATHLEN]; // PATHLEN is defined to 256 3989 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3990 * ctx->di); 3991 * 3992 * // Consume buf, for example push it to 3993 * // userspace via bpf_perf_event_output(); we 3994 * // can use res (the string length) as event 3995 * // size, after checking its boundaries. 3996 * } 3997 * 3998 * In comparison, using **bpf_probe_read_user**\ () helper here 3999 * instead to read the string would require to estimate the length 4000 * at compile time, and would often result in copying more memory 4001 * than necessary. 4002 * 4003 * Another useful use case is when parsing individual process 4004 * arguments or individual environment variables navigating 4005 * *current*\ **->mm->arg_start** and *current*\ 4006 * **->mm->env_start**: using this helper and the return value, 4007 * one can quickly iterate at the right offset of the memory area. 4008 * Return 4009 * On success, the strictly positive length of the output string, 4010 * including the trailing NUL character. On error, a negative 4011 * value. 4012 * 4013 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4014 * Description 4015 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4016 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4017 * Return 4018 * On success, the strictly positive length of the string, including 4019 * the trailing NUL character. On error, a negative value. 4020 * 4021 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4022 * Description 4023 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4024 * *rcv_nxt* is the ack_seq to be sent out. 4025 * Return 4026 * 0 on success, or a negative error in case of failure. 4027 * 4028 * long bpf_send_signal_thread(u32 sig) 4029 * Description 4030 * Send signal *sig* to the thread corresponding to the current task. 4031 * Return 4032 * 0 on success or successfully queued. 4033 * 4034 * **-EBUSY** if work queue under nmi is full. 4035 * 4036 * **-EINVAL** if *sig* is invalid. 4037 * 4038 * **-EPERM** if no permission to send the *sig*. 4039 * 4040 * **-EAGAIN** if bpf program can try again. 4041 * 4042 * u64 bpf_jiffies64(void) 4043 * Description 4044 * Obtain the 64bit jiffies 4045 * Return 4046 * The 64 bit jiffies 4047 * 4048 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4049 * Description 4050 * For an eBPF program attached to a perf event, retrieve the 4051 * branch records (**struct perf_branch_entry**) associated to *ctx* 4052 * and store it in the buffer pointed by *buf* up to size 4053 * *size* bytes. 4054 * Return 4055 * On success, number of bytes written to *buf*. On error, a 4056 * negative value. 4057 * 4058 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4059 * instead return the number of bytes required to store all the 4060 * branch entries. If this flag is set, *buf* may be NULL. 4061 * 4062 * **-EINVAL** if arguments invalid or **size** not a multiple 4063 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4064 * 4065 * **-ENOENT** if architecture does not support branch records. 4066 * 4067 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4068 * Description 4069 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4070 * *namespace* will be returned in *nsdata*. 4071 * Return 4072 * 0 on success, or one of the following in case of failure: 4073 * 4074 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4075 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4076 * 4077 * **-ENOENT** if pidns does not exists for the current task. 4078 * 4079 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4080 * Description 4081 * Write raw *data* blob into a special BPF perf event held by 4082 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4083 * event must have the following attributes: **PERF_SAMPLE_RAW** 4084 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4085 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4086 * 4087 * The *flags* are used to indicate the index in *map* for which 4088 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4089 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4090 * to indicate that the index of the current CPU core should be 4091 * used. 4092 * 4093 * The value to write, of *size*, is passed through eBPF stack and 4094 * pointed by *data*. 4095 * 4096 * *ctx* is a pointer to in-kernel struct xdp_buff. 4097 * 4098 * This helper is similar to **bpf_perf_eventoutput**\ () but 4099 * restricted to raw_tracepoint bpf programs. 4100 * Return 4101 * 0 on success, or a negative error in case of failure. 4102 * 4103 * u64 bpf_get_netns_cookie(void *ctx) 4104 * Description 4105 * Retrieve the cookie (generated by the kernel) of the network 4106 * namespace the input *ctx* is associated with. The network 4107 * namespace cookie remains stable for its lifetime and provides 4108 * a global identifier that can be assumed unique. If *ctx* is 4109 * NULL, then the helper returns the cookie for the initial 4110 * network namespace. The cookie itself is very similar to that 4111 * of **bpf_get_socket_cookie**\ () helper, but for network 4112 * namespaces instead of sockets. 4113 * Return 4114 * A 8-byte long opaque number. 4115 * 4116 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4117 * Description 4118 * Return id of cgroup v2 that is ancestor of the cgroup associated 4119 * with the current task at the *ancestor_level*. The root cgroup 4120 * is at *ancestor_level* zero and each step down the hierarchy 4121 * increments the level. If *ancestor_level* == level of cgroup 4122 * associated with the current task, then return value will be the 4123 * same as that of **bpf_get_current_cgroup_id**\ (). 4124 * 4125 * The helper is useful to implement policies based on cgroups 4126 * that are upper in hierarchy than immediate cgroup associated 4127 * with the current task. 4128 * 4129 * The format of returned id and helper limitations are same as in 4130 * **bpf_get_current_cgroup_id**\ (). 4131 * Return 4132 * The id is returned or 0 in case the id could not be retrieved. 4133 * 4134 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4135 * Description 4136 * Helper is overloaded depending on BPF program type. This 4137 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4138 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4139 * 4140 * Assign the *sk* to the *skb*. When combined with appropriate 4141 * routing configuration to receive the packet towards the socket, 4142 * will cause *skb* to be delivered to the specified socket. 4143 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4144 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4145 * interfere with successful delivery to the socket. 4146 * 4147 * This operation is only valid from TC ingress path. 4148 * 4149 * The *flags* argument must be zero. 4150 * Return 4151 * 0 on success, or a negative error in case of failure: 4152 * 4153 * **-EINVAL** if specified *flags* are not supported. 4154 * 4155 * **-ENOENT** if the socket is unavailable for assignment. 4156 * 4157 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4158 * 4159 * **-EOPNOTSUPP** if the operation is not supported, for example 4160 * a call from outside of TC ingress. 4161 * 4162 * **-ESOCKTNOSUPPORT** if the socket type is not supported 4163 * (reuseport). 4164 * 4165 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4166 * Description 4167 * Helper is overloaded depending on BPF program type. This 4168 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4169 * 4170 * Select the *sk* as a result of a socket lookup. 4171 * 4172 * For the operation to succeed passed socket must be compatible 4173 * with the packet description provided by the *ctx* object. 4174 * 4175 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4176 * be an exact match. While IP family (**AF_INET** or 4177 * **AF_INET6**) must be compatible, that is IPv6 sockets 4178 * that are not v6-only can be selected for IPv4 packets. 4179 * 4180 * Only TCP listeners and UDP unconnected sockets can be 4181 * selected. *sk* can also be NULL to reset any previous 4182 * selection. 4183 * 4184 * *flags* argument can combination of following values: 4185 * 4186 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4187 * socket selection, potentially done by a BPF program 4188 * that ran before us. 4189 * 4190 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4191 * load-balancing within reuseport group for the socket 4192 * being selected. 4193 * 4194 * On success *ctx->sk* will point to the selected socket. 4195 * 4196 * Return 4197 * 0 on success, or a negative errno in case of failure. 4198 * 4199 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4200 * not compatible with packet family (*ctx->family*). 4201 * 4202 * * **-EEXIST** if socket has been already selected, 4203 * potentially by another program, and 4204 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4205 * 4206 * * **-EINVAL** if unsupported flags were specified. 4207 * 4208 * * **-EPROTOTYPE** if socket L4 protocol 4209 * (*sk->protocol*) doesn't match packet protocol 4210 * (*ctx->protocol*). 4211 * 4212 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4213 * state (TCP listening or UDP unconnected). 4214 * 4215 * u64 bpf_ktime_get_boot_ns(void) 4216 * Description 4217 * Return the time elapsed since system boot, in nanoseconds. 4218 * Does include the time the system was suspended. 4219 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4220 * Return 4221 * Current *ktime*. 4222 * 4223 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4224 * Description 4225 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4226 * out the format string. 4227 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4228 * the format string itself. The *data* and *data_len* are format string 4229 * arguments. The *data* are a **u64** array and corresponding format string 4230 * values are stored in the array. For strings and pointers where pointees 4231 * are accessed, only the pointer values are stored in the *data* array. 4232 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4233 * 4234 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4235 * Reading kernel memory may fail due to either invalid address or 4236 * valid address but requiring a major memory fault. If reading kernel memory 4237 * fails, the string for **%s** will be an empty string, and the ip 4238 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4239 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4240 * Return 4241 * 0 on success, or a negative error in case of failure: 4242 * 4243 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4244 * by returning 1 from bpf program. 4245 * 4246 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4247 * 4248 * **-E2BIG** if *fmt* contains too many format specifiers. 4249 * 4250 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4251 * 4252 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4253 * Description 4254 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4255 * The *m* represents the seq_file. The *data* and *len* represent the 4256 * data to write in bytes. 4257 * Return 4258 * 0 on success, or a negative error in case of failure: 4259 * 4260 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4261 * 4262 * u64 bpf_sk_cgroup_id(void *sk) 4263 * Description 4264 * Return the cgroup v2 id of the socket *sk*. 4265 * 4266 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4267 * returned from **bpf_sk_lookup_xxx**\ (), 4268 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4269 * same as in **bpf_skb_cgroup_id**\ (). 4270 * 4271 * This helper is available only if the kernel was compiled with 4272 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4273 * Return 4274 * The id is returned or 0 in case the id could not be retrieved. 4275 * 4276 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4277 * Description 4278 * Return id of cgroup v2 that is ancestor of cgroup associated 4279 * with the *sk* at the *ancestor_level*. The root cgroup is at 4280 * *ancestor_level* zero and each step down the hierarchy 4281 * increments the level. If *ancestor_level* == level of cgroup 4282 * associated with *sk*, then return value will be same as that 4283 * of **bpf_sk_cgroup_id**\ (). 4284 * 4285 * The helper is useful to implement policies based on cgroups 4286 * that are upper in hierarchy than immediate cgroup associated 4287 * with *sk*. 4288 * 4289 * The format of returned id and helper limitations are same as in 4290 * **bpf_sk_cgroup_id**\ (). 4291 * Return 4292 * The id is returned or 0 in case the id could not be retrieved. 4293 * 4294 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4295 * Description 4296 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4297 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4298 * of new data availability is sent. 4299 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4300 * of new data availability is sent unconditionally. 4301 * If **0** is specified in *flags*, an adaptive notification 4302 * of new data availability is sent. 4303 * 4304 * An adaptive notification is a notification sent whenever the user-space 4305 * process has caught up and consumed all available payloads. In case the user-space 4306 * process is still processing a previous payload, then no notification is needed 4307 * as it will process the newly added payload automatically. 4308 * Return 4309 * 0 on success, or a negative error in case of failure. 4310 * 4311 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4312 * Description 4313 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4314 * *flags* must be 0. 4315 * Return 4316 * Valid pointer with *size* bytes of memory available; NULL, 4317 * otherwise. 4318 * 4319 * void bpf_ringbuf_submit(void *data, u64 flags) 4320 * Description 4321 * Submit reserved ring buffer sample, pointed to by *data*. 4322 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4323 * of new data availability is sent. 4324 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4325 * of new data availability is sent unconditionally. 4326 * If **0** is specified in *flags*, an adaptive notification 4327 * of new data availability is sent. 4328 * 4329 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4330 * Return 4331 * Nothing. Always succeeds. 4332 * 4333 * void bpf_ringbuf_discard(void *data, u64 flags) 4334 * Description 4335 * Discard reserved ring buffer sample, pointed to by *data*. 4336 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4337 * of new data availability is sent. 4338 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4339 * of new data availability is sent unconditionally. 4340 * If **0** is specified in *flags*, an adaptive notification 4341 * of new data availability is sent. 4342 * 4343 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4344 * Return 4345 * Nothing. Always succeeds. 4346 * 4347 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4348 * Description 4349 * Query various characteristics of provided ring buffer. What 4350 * exactly is queries is determined by *flags*: 4351 * 4352 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4353 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4354 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4355 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4356 * 4357 * Data returned is just a momentary snapshot of actual values 4358 * and could be inaccurate, so this facility should be used to 4359 * power heuristics and for reporting, not to make 100% correct 4360 * calculation. 4361 * Return 4362 * Requested value, or 0, if *flags* are not recognized. 4363 * 4364 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4365 * Description 4366 * Change the skbs checksum level by one layer up or down, or 4367 * reset it entirely to none in order to have the stack perform 4368 * checksum validation. The level is applicable to the following 4369 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4370 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4371 * through **bpf_skb_adjust_room**\ () helper with passing in 4372 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4373 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4374 * the UDP header is removed. Similarly, an encap of the latter 4375 * into the former could be accompanied by a helper call to 4376 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4377 * skb is still intended to be processed in higher layers of the 4378 * stack instead of just egressing at tc. 4379 * 4380 * There are three supported level settings at this time: 4381 * 4382 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4383 * with CHECKSUM_UNNECESSARY. 4384 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4385 * with CHECKSUM_UNNECESSARY. 4386 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4387 * sets CHECKSUM_NONE to force checksum validation by the stack. 4388 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4389 * skb->csum_level. 4390 * Return 4391 * 0 on success, or a negative error in case of failure. In the 4392 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4393 * is returned or the error code -EACCES in case the skb is not 4394 * subject to CHECKSUM_UNNECESSARY. 4395 * 4396 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4397 * Description 4398 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4399 * Return 4400 * *sk* if casting is valid, or **NULL** otherwise. 4401 * 4402 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4403 * Description 4404 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4405 * Return 4406 * *sk* if casting is valid, or **NULL** otherwise. 4407 * 4408 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4409 * Description 4410 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4411 * Return 4412 * *sk* if casting is valid, or **NULL** otherwise. 4413 * 4414 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4415 * Description 4416 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4417 * Return 4418 * *sk* if casting is valid, or **NULL** otherwise. 4419 * 4420 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4421 * Description 4422 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4423 * Return 4424 * *sk* if casting is valid, or **NULL** otherwise. 4425 * 4426 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4427 * Description 4428 * Return a user or a kernel stack in bpf program provided buffer. 4429 * To achieve this, the helper needs *task*, which is a valid 4430 * pointer to **struct task_struct**. To store the stacktrace, the 4431 * bpf program provides *buf* with a nonnegative *size*. 4432 * 4433 * The last argument, *flags*, holds the number of stack frames to 4434 * skip (from 0 to 255), masked with 4435 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4436 * the following flags: 4437 * 4438 * **BPF_F_USER_STACK** 4439 * Collect a user space stack instead of a kernel stack. 4440 * **BPF_F_USER_BUILD_ID** 4441 * Collect buildid+offset instead of ips for user stack, 4442 * only valid if **BPF_F_USER_STACK** is also specified. 4443 * 4444 * **bpf_get_task_stack**\ () can collect up to 4445 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4446 * to sufficient large buffer size. Note that 4447 * this limit can be controlled with the **sysctl** program, and 4448 * that it should be manually increased in order to profile long 4449 * user stacks (such as stacks for Java programs). To do so, use: 4450 * 4451 * :: 4452 * 4453 * # sysctl kernel.perf_event_max_stack=<new value> 4454 * Return 4455 * The non-negative copied *buf* length equal to or less than 4456 * *size* on success, or a negative error in case of failure. 4457 * 4458 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4459 * Description 4460 * Load header option. Support reading a particular TCP header 4461 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4462 * 4463 * If *flags* is 0, it will search the option from the 4464 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4465 * has details on what skb_data contains under different 4466 * *skops*\ **->op**. 4467 * 4468 * The first byte of the *searchby_res* specifies the 4469 * kind that it wants to search. 4470 * 4471 * If the searching kind is an experimental kind 4472 * (i.e. 253 or 254 according to RFC6994). It also 4473 * needs to specify the "magic" which is either 4474 * 2 bytes or 4 bytes. It then also needs to 4475 * specify the size of the magic by using 4476 * the 2nd byte which is "kind-length" of a TCP 4477 * header option and the "kind-length" also 4478 * includes the first 2 bytes "kind" and "kind-length" 4479 * itself as a normal TCP header option also does. 4480 * 4481 * For example, to search experimental kind 254 with 4482 * 2 byte magic 0xeB9F, the searchby_res should be 4483 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4484 * 4485 * To search for the standard window scale option (3), 4486 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4487 * Note, kind-length must be 0 for regular option. 4488 * 4489 * Searching for No-Op (0) and End-of-Option-List (1) are 4490 * not supported. 4491 * 4492 * *len* must be at least 2 bytes which is the minimal size 4493 * of a header option. 4494 * 4495 * Supported flags: 4496 * 4497 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4498 * saved_syn packet or the just-received syn packet. 4499 * 4500 * Return 4501 * > 0 when found, the header option is copied to *searchby_res*. 4502 * The return value is the total length copied. On failure, a 4503 * negative error code is returned: 4504 * 4505 * **-EINVAL** if a parameter is invalid. 4506 * 4507 * **-ENOMSG** if the option is not found. 4508 * 4509 * **-ENOENT** if no syn packet is available when 4510 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4511 * 4512 * **-ENOSPC** if there is not enough space. Only *len* number of 4513 * bytes are copied. 4514 * 4515 * **-EFAULT** on failure to parse the header options in the 4516 * packet. 4517 * 4518 * **-EPERM** if the helper cannot be used under the current 4519 * *skops*\ **->op**. 4520 * 4521 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4522 * Description 4523 * Store header option. The data will be copied 4524 * from buffer *from* with length *len* to the TCP header. 4525 * 4526 * The buffer *from* should have the whole option that 4527 * includes the kind, kind-length, and the actual 4528 * option data. The *len* must be at least kind-length 4529 * long. The kind-length does not have to be 4 byte 4530 * aligned. The kernel will take care of the padding 4531 * and setting the 4 bytes aligned value to th->doff. 4532 * 4533 * This helper will check for duplicated option 4534 * by searching the same option in the outgoing skb. 4535 * 4536 * This helper can only be called during 4537 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4538 * 4539 * Return 4540 * 0 on success, or negative error in case of failure: 4541 * 4542 * **-EINVAL** If param is invalid. 4543 * 4544 * **-ENOSPC** if there is not enough space in the header. 4545 * Nothing has been written 4546 * 4547 * **-EEXIST** if the option already exists. 4548 * 4549 * **-EFAULT** on failure to parse the existing header options. 4550 * 4551 * **-EPERM** if the helper cannot be used under the current 4552 * *skops*\ **->op**. 4553 * 4554 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4555 * Description 4556 * Reserve *len* bytes for the bpf header option. The 4557 * space will be used by **bpf_store_hdr_opt**\ () later in 4558 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4559 * 4560 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4561 * the total number of bytes will be reserved. 4562 * 4563 * This helper can only be called during 4564 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4565 * 4566 * Return 4567 * 0 on success, or negative error in case of failure: 4568 * 4569 * **-EINVAL** if a parameter is invalid. 4570 * 4571 * **-ENOSPC** if there is not enough space in the header. 4572 * 4573 * **-EPERM** if the helper cannot be used under the current 4574 * *skops*\ **->op**. 4575 * 4576 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4577 * Description 4578 * Get a bpf_local_storage from an *inode*. 4579 * 4580 * Logically, it could be thought of as getting the value from 4581 * a *map* with *inode* as the **key**. From this 4582 * perspective, the usage is not much different from 4583 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4584 * helper enforces the key must be an inode and the map must also 4585 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4586 * 4587 * Underneath, the value is stored locally at *inode* instead of 4588 * the *map*. The *map* is used as the bpf-local-storage 4589 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4590 * searched against all bpf_local_storage residing at *inode*. 4591 * 4592 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4593 * used such that a new bpf_local_storage will be 4594 * created if one does not exist. *value* can be used 4595 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4596 * the initial value of a bpf_local_storage. If *value* is 4597 * **NULL**, the new bpf_local_storage will be zero initialized. 4598 * Return 4599 * A bpf_local_storage pointer is returned on success. 4600 * 4601 * **NULL** if not found or there was an error in adding 4602 * a new bpf_local_storage. 4603 * 4604 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4605 * Description 4606 * Delete a bpf_local_storage from an *inode*. 4607 * Return 4608 * 0 on success. 4609 * 4610 * **-ENOENT** if the bpf_local_storage cannot be found. 4611 * 4612 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4613 * Description 4614 * Return full path for given **struct path** object, which 4615 * needs to be the kernel BTF *path* object. The path is 4616 * returned in the provided buffer *buf* of size *sz* and 4617 * is zero terminated. 4618 * 4619 * Return 4620 * On success, the strictly positive length of the string, 4621 * including the trailing NUL character. On error, a negative 4622 * value. 4623 * 4624 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4625 * Description 4626 * Read *size* bytes from user space address *user_ptr* and store 4627 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4628 * Return 4629 * 0 on success, or a negative error in case of failure. 4630 * 4631 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4632 * Description 4633 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4634 * using *ptr*->type_id. This value should specify the type 4635 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4636 * can be used to look up vmlinux BTF type ids. Traversing the 4637 * data structure using BTF, the type information and values are 4638 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4639 * the pointer data is carried out to avoid kernel crashes during 4640 * operation. Smaller types can use string space on the stack; 4641 * larger programs can use map data to store the string 4642 * representation. 4643 * 4644 * The string can be subsequently shared with userspace via 4645 * bpf_perf_event_output() or ring buffer interfaces. 4646 * bpf_trace_printk() is to be avoided as it places too small 4647 * a limit on string size to be useful. 4648 * 4649 * *flags* is a combination of 4650 * 4651 * **BTF_F_COMPACT** 4652 * no formatting around type information 4653 * **BTF_F_NONAME** 4654 * no struct/union member names/types 4655 * **BTF_F_PTR_RAW** 4656 * show raw (unobfuscated) pointer values; 4657 * equivalent to printk specifier %px. 4658 * **BTF_F_ZERO** 4659 * show zero-valued struct/union members; they 4660 * are not displayed by default 4661 * 4662 * Return 4663 * The number of bytes that were written (or would have been 4664 * written if output had to be truncated due to string size), 4665 * or a negative error in cases of failure. 4666 * 4667 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4668 * Description 4669 * Use BTF to write to seq_write a string representation of 4670 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4671 * *flags* are identical to those used for bpf_snprintf_btf. 4672 * Return 4673 * 0 on success or a negative error in case of failure. 4674 * 4675 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4676 * Description 4677 * See **bpf_get_cgroup_classid**\ () for the main description. 4678 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4679 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4680 * associated socket instead of the current process. 4681 * Return 4682 * The id is returned or 0 in case the id could not be retrieved. 4683 * 4684 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4685 * Description 4686 * Redirect the packet to another net device of index *ifindex* 4687 * and fill in L2 addresses from neighboring subsystem. This helper 4688 * is somewhat similar to **bpf_redirect**\ (), except that it 4689 * populates L2 addresses as well, meaning, internally, the helper 4690 * relies on the neighbor lookup for the L2 address of the nexthop. 4691 * 4692 * The helper will perform a FIB lookup based on the skb's 4693 * networking header to get the address of the next hop, unless 4694 * this is supplied by the caller in the *params* argument. The 4695 * *plen* argument indicates the len of *params* and should be set 4696 * to 0 if *params* is NULL. 4697 * 4698 * The *flags* argument is reserved and must be 0. The helper is 4699 * currently only supported for tc BPF program types, and enabled 4700 * for IPv4 and IPv6 protocols. 4701 * Return 4702 * The helper returns **TC_ACT_REDIRECT** on success or 4703 * **TC_ACT_SHOT** on error. 4704 * 4705 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4706 * Description 4707 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4708 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4709 * extern variable decorated with '__ksym'. For ksym, there is a 4710 * global var (either static or global) defined of the same name 4711 * in the kernel. The ksym is percpu if the global var is percpu. 4712 * The returned pointer points to the global percpu var on *cpu*. 4713 * 4714 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4715 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4716 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4717 * bpf_per_cpu_ptr() must check the returned value. 4718 * Return 4719 * A pointer pointing to the kernel percpu variable on *cpu*, or 4720 * NULL, if *cpu* is invalid. 4721 * 4722 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4723 * Description 4724 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4725 * pointer to the percpu kernel variable on this cpu. See the 4726 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4727 * 4728 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4729 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4730 * never return NULL. 4731 * Return 4732 * A pointer pointing to the kernel percpu variable on this cpu. 4733 * 4734 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4735 * Description 4736 * Redirect the packet to another net device of index *ifindex*. 4737 * This helper is somewhat similar to **bpf_redirect**\ (), except 4738 * that the redirection happens to the *ifindex*' peer device and 4739 * the netns switch takes place from ingress to ingress without 4740 * going through the CPU's backlog queue. 4741 * 4742 * The *flags* argument is reserved and must be 0. The helper is 4743 * currently only supported for tc BPF program types at the ingress 4744 * hook and for veth device types. The peer device must reside in a 4745 * different network namespace. 4746 * Return 4747 * The helper returns **TC_ACT_REDIRECT** on success or 4748 * **TC_ACT_SHOT** on error. 4749 * 4750 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4751 * Description 4752 * Get a bpf_local_storage from the *task*. 4753 * 4754 * Logically, it could be thought of as getting the value from 4755 * a *map* with *task* as the **key**. From this 4756 * perspective, the usage is not much different from 4757 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4758 * helper enforces the key must be a task_struct and the map must also 4759 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4760 * 4761 * Underneath, the value is stored locally at *task* instead of 4762 * the *map*. The *map* is used as the bpf-local-storage 4763 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4764 * searched against all bpf_local_storage residing at *task*. 4765 * 4766 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4767 * used such that a new bpf_local_storage will be 4768 * created if one does not exist. *value* can be used 4769 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4770 * the initial value of a bpf_local_storage. If *value* is 4771 * **NULL**, the new bpf_local_storage will be zero initialized. 4772 * Return 4773 * A bpf_local_storage pointer is returned on success. 4774 * 4775 * **NULL** if not found or there was an error in adding 4776 * a new bpf_local_storage. 4777 * 4778 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4779 * Description 4780 * Delete a bpf_local_storage from a *task*. 4781 * Return 4782 * 0 on success. 4783 * 4784 * **-ENOENT** if the bpf_local_storage cannot be found. 4785 * 4786 * struct task_struct *bpf_get_current_task_btf(void) 4787 * Description 4788 * Return a BTF pointer to the "current" task. 4789 * This pointer can also be used in helpers that accept an 4790 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4791 * Return 4792 * Pointer to the current task. 4793 * 4794 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4795 * Description 4796 * Set or clear certain options on *bprm*: 4797 * 4798 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4799 * which sets the **AT_SECURE** auxv for glibc. The bit 4800 * is cleared if the flag is not specified. 4801 * Return 4802 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4803 * 4804 * u64 bpf_ktime_get_coarse_ns(void) 4805 * Description 4806 * Return a coarse-grained version of the time elapsed since 4807 * system boot, in nanoseconds. Does not include time the system 4808 * was suspended. 4809 * 4810 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4811 * Return 4812 * Current *ktime*. 4813 * 4814 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4815 * Description 4816 * Returns the stored IMA hash of the *inode* (if it's available). 4817 * If the hash is larger than *size*, then only *size* 4818 * bytes will be copied to *dst* 4819 * Return 4820 * The **hash_algo** is returned on success, 4821 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4822 * invalid arguments are passed. 4823 * 4824 * struct socket *bpf_sock_from_file(struct file *file) 4825 * Description 4826 * If the given file represents a socket, returns the associated 4827 * socket. 4828 * Return 4829 * A pointer to a struct socket on success or NULL if the file is 4830 * not a socket. 4831 * 4832 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4833 * Description 4834 * Check packet size against exceeding MTU of net device (based 4835 * on *ifindex*). This helper will likely be used in combination 4836 * with helpers that adjust/change the packet size. 4837 * 4838 * The argument *len_diff* can be used for querying with a planned 4839 * size change. This allows to check MTU prior to changing packet 4840 * ctx. Providing a *len_diff* adjustment that is larger than the 4841 * actual packet size (resulting in negative packet size) will in 4842 * principle not exceed the MTU, which is why it is not considered 4843 * a failure. Other BPF helpers are needed for performing the 4844 * planned size change; therefore the responsibility for catching 4845 * a negative packet size belongs in those helpers. 4846 * 4847 * Specifying *ifindex* zero means the MTU check is performed 4848 * against the current net device. This is practical if this isn't 4849 * used prior to redirect. 4850 * 4851 * On input *mtu_len* must be a valid pointer, else verifier will 4852 * reject BPF program. If the value *mtu_len* is initialized to 4853 * zero then the ctx packet size is use. When value *mtu_len* is 4854 * provided as input this specify the L3 length that the MTU check 4855 * is done against. Remember XDP and TC length operate at L2, but 4856 * this value is L3 as this correlate to MTU and IP-header tot_len 4857 * values which are L3 (similar behavior as bpf_fib_lookup). 4858 * 4859 * The Linux kernel route table can configure MTUs on a more 4860 * specific per route level, which is not provided by this helper. 4861 * For route level MTU checks use the **bpf_fib_lookup**\ () 4862 * helper. 4863 * 4864 * *ctx* is either **struct xdp_md** for XDP programs or 4865 * **struct sk_buff** for tc cls_act programs. 4866 * 4867 * The *flags* argument can be a combination of one or more of the 4868 * following values: 4869 * 4870 * **BPF_MTU_CHK_SEGS** 4871 * This flag will only works for *ctx* **struct sk_buff**. 4872 * If packet context contains extra packet segment buffers 4873 * (often knows as GSO skb), then MTU check is harder to 4874 * check at this point, because in transmit path it is 4875 * possible for the skb packet to get re-segmented 4876 * (depending on net device features). This could still be 4877 * a MTU violation, so this flag enables performing MTU 4878 * check against segments, with a different violation 4879 * return code to tell it apart. Check cannot use len_diff. 4880 * 4881 * On return *mtu_len* pointer contains the MTU value of the net 4882 * device. Remember the net device configured MTU is the L3 size, 4883 * which is returned here and XDP and TC length operate at L2. 4884 * Helper take this into account for you, but remember when using 4885 * MTU value in your BPF-code. 4886 * 4887 * Return 4888 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4889 * 4890 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4891 * 4892 * MTU violations return positive values, but also populate MTU 4893 * value in *mtu_len* pointer, as this can be needed for 4894 * implementing PMTU handing: 4895 * 4896 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4897 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4898 * 4899 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4900 * Description 4901 * For each element in **map**, call **callback_fn** function with 4902 * **map**, **callback_ctx** and other map-specific parameters. 4903 * The **callback_fn** should be a static function and 4904 * the **callback_ctx** should be a pointer to the stack. 4905 * The **flags** is used to control certain aspects of the helper. 4906 * Currently, the **flags** must be 0. 4907 * 4908 * The following are a list of supported map types and their 4909 * respective expected callback signatures: 4910 * 4911 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4912 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4913 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4914 * 4915 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4916 * 4917 * For per_cpu maps, the map_value is the value on the cpu where the 4918 * bpf_prog is running. 4919 * 4920 * If **callback_fn** return 0, the helper will continue to the next 4921 * element. If return value is 1, the helper will skip the rest of 4922 * elements and return. Other return values are not used now. 4923 * 4924 * Return 4925 * The number of traversed map elements for success, **-EINVAL** for 4926 * invalid **flags**. 4927 * 4928 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4929 * Description 4930 * Outputs a string into the **str** buffer of size **str_size** 4931 * based on a format string stored in a read-only map pointed by 4932 * **fmt**. 4933 * 4934 * Each format specifier in **fmt** corresponds to one u64 element 4935 * in the **data** array. For strings and pointers where pointees 4936 * are accessed, only the pointer values are stored in the *data* 4937 * array. The *data_len* is the size of *data* in bytes - must be 4938 * a multiple of 8. 4939 * 4940 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4941 * memory. Reading kernel memory may fail due to either invalid 4942 * address or valid address but requiring a major memory fault. If 4943 * reading kernel memory fails, the string for **%s** will be an 4944 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4945 * Not returning error to bpf program is consistent with what 4946 * **bpf_trace_printk**\ () does for now. 4947 * 4948 * Return 4949 * The strictly positive length of the formatted string, including 4950 * the trailing zero character. If the return value is greater than 4951 * **str_size**, **str** contains a truncated string, guaranteed to 4952 * be zero-terminated except when **str_size** is 0. 4953 * 4954 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4955 * 4956 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4957 * Description 4958 * Execute bpf syscall with given arguments. 4959 * Return 4960 * A syscall result. 4961 * 4962 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4963 * Description 4964 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4965 * Return 4966 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4967 * 4968 * long bpf_sys_close(u32 fd) 4969 * Description 4970 * Execute close syscall for given FD. 4971 * Return 4972 * A syscall result. 4973 * 4974 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4975 * Description 4976 * Initialize the timer. 4977 * First 4 bits of *flags* specify clockid. 4978 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4979 * All other bits of *flags* are reserved. 4980 * The verifier will reject the program if *timer* is not from 4981 * the same *map*. 4982 * Return 4983 * 0 on success. 4984 * **-EBUSY** if *timer* is already initialized. 4985 * **-EINVAL** if invalid *flags* are passed. 4986 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4987 * The user space should either hold a file descriptor to a map with timers 4988 * or pin such map in bpffs. When map is unpinned or file descriptor is 4989 * closed all timers in the map will be cancelled and freed. 4990 * 4991 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4992 * Description 4993 * Configure the timer to call *callback_fn* static function. 4994 * Return 4995 * 0 on success. 4996 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4997 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4998 * The user space should either hold a file descriptor to a map with timers 4999 * or pin such map in bpffs. When map is unpinned or file descriptor is 5000 * closed all timers in the map will be cancelled and freed. 5001 * 5002 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5003 * Description 5004 * Set timer expiration N nanoseconds from the current time. The 5005 * configured callback will be invoked in soft irq context on some cpu 5006 * and will not repeat unless another bpf_timer_start() is made. 5007 * In such case the next invocation can migrate to a different cpu. 5008 * Since struct bpf_timer is a field inside map element the map 5009 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5010 * of BPF program to make sure that callback_fn code stays valid. 5011 * When user space reference to a map reaches zero all timers 5012 * in a map are cancelled and corresponding program's refcnts are 5013 * decremented. This is done to make sure that Ctrl-C of a user 5014 * process doesn't leave any timers running. If map is pinned in 5015 * bpffs the callback_fn can re-arm itself indefinitely. 5016 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5017 * cancel and free the timer in the given map element. 5018 * The map can contain timers that invoke callback_fn-s from different 5019 * programs. The same callback_fn can serve different timers from 5020 * different maps if key/value layout matches across maps. 5021 * Every bpf_timer_set_callback() can have different callback_fn. 5022 * 5023 * *flags* can be one of: 5024 * 5025 * **BPF_F_TIMER_ABS** 5026 * Start the timer in absolute expire value instead of the 5027 * default relative one. 5028 * 5029 * Return 5030 * 0 on success. 5031 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5032 * or invalid *flags* are passed. 5033 * 5034 * long bpf_timer_cancel(struct bpf_timer *timer) 5035 * Description 5036 * Cancel the timer and wait for callback_fn to finish if it was running. 5037 * Return 5038 * 0 if the timer was not active. 5039 * 1 if the timer was active. 5040 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5041 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5042 * own timer which would have led to a deadlock otherwise. 5043 * 5044 * u64 bpf_get_func_ip(void *ctx) 5045 * Description 5046 * Get address of the traced function (for tracing and kprobe programs). 5047 * Return 5048 * Address of the traced function. 5049 * 0 for kprobes placed within the function (not at the entry). 5050 * 5051 * u64 bpf_get_attach_cookie(void *ctx) 5052 * Description 5053 * Get bpf_cookie value provided (optionally) during the program 5054 * attachment. It might be different for each individual 5055 * attachment, even if BPF program itself is the same. 5056 * Expects BPF program context *ctx* as a first argument. 5057 * 5058 * Supported for the following program types: 5059 * - kprobe/uprobe; 5060 * - tracepoint; 5061 * - perf_event. 5062 * Return 5063 * Value specified by user at BPF link creation/attachment time 5064 * or 0, if it was not specified. 5065 * 5066 * long bpf_task_pt_regs(struct task_struct *task) 5067 * Description 5068 * Get the struct pt_regs associated with **task**. 5069 * Return 5070 * A pointer to struct pt_regs. 5071 * 5072 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5073 * Description 5074 * Get branch trace from hardware engines like Intel LBR. The 5075 * hardware engine is stopped shortly after the helper is 5076 * called. Therefore, the user need to filter branch entries 5077 * based on the actual use case. To capture branch trace 5078 * before the trigger point of the BPF program, the helper 5079 * should be called at the beginning of the BPF program. 5080 * 5081 * The data is stored as struct perf_branch_entry into output 5082 * buffer *entries*. *size* is the size of *entries* in bytes. 5083 * *flags* is reserved for now and must be zero. 5084 * 5085 * Return 5086 * On success, number of bytes written to *buf*. On error, a 5087 * negative value. 5088 * 5089 * **-EINVAL** if *flags* is not zero. 5090 * 5091 * **-ENOENT** if architecture does not support branch records. 5092 * 5093 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5094 * Description 5095 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5096 * to format and can handle more format args as a result. 5097 * 5098 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5099 * Return 5100 * The number of bytes written to the buffer, or a negative error 5101 * in case of failure. 5102 * 5103 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5104 * Description 5105 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5106 * Return 5107 * *sk* if casting is valid, or **NULL** otherwise. 5108 * 5109 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5110 * Description 5111 * Get the address of a kernel symbol, returned in *res*. *res* is 5112 * set to 0 if the symbol is not found. 5113 * Return 5114 * On success, zero. On error, a negative value. 5115 * 5116 * **-EINVAL** if *flags* is not zero. 5117 * 5118 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5119 * 5120 * **-ENOENT** if symbol is not found. 5121 * 5122 * **-EPERM** if caller does not have permission to obtain kernel address. 5123 * 5124 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5125 * Description 5126 * Find vma of *task* that contains *addr*, call *callback_fn* 5127 * function with *task*, *vma*, and *callback_ctx*. 5128 * The *callback_fn* should be a static function and 5129 * the *callback_ctx* should be a pointer to the stack. 5130 * The *flags* is used to control certain aspects of the helper. 5131 * Currently, the *flags* must be 0. 5132 * 5133 * The expected callback signature is 5134 * 5135 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5136 * 5137 * Return 5138 * 0 on success. 5139 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5140 * **-EBUSY** if failed to try lock mmap_lock. 5141 * **-EINVAL** for invalid **flags**. 5142 * 5143 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5144 * Description 5145 * For **nr_loops**, call **callback_fn** function 5146 * with **callback_ctx** as the context parameter. 5147 * The **callback_fn** should be a static function and 5148 * the **callback_ctx** should be a pointer to the stack. 5149 * The **flags** is used to control certain aspects of the helper. 5150 * Currently, the **flags** must be 0. Currently, nr_loops is 5151 * limited to 1 << 23 (~8 million) loops. 5152 * 5153 * long (\*callback_fn)(u32 index, void \*ctx); 5154 * 5155 * where **index** is the current index in the loop. The index 5156 * is zero-indexed. 5157 * 5158 * If **callback_fn** returns 0, the helper will continue to the next 5159 * loop. If return value is 1, the helper will skip the rest of 5160 * the loops and return. Other return values are not used now, 5161 * and will be rejected by the verifier. 5162 * 5163 * Return 5164 * The number of loops performed, **-EINVAL** for invalid **flags**, 5165 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5166 * 5167 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5168 * Description 5169 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5170 * to be null-terminated and **s1_sz** is the maximum storage 5171 * size of **s1**. **s2** must be a read-only string. 5172 * Return 5173 * An integer less than, equal to, or greater than zero 5174 * if the first **s1_sz** bytes of **s1** is found to be 5175 * less than, to match, or be greater than **s2**. 5176 * 5177 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5178 * Description 5179 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5180 * returned in **value**. 5181 * 5182 * Return 5183 * 0 on success. 5184 * **-EINVAL** if n >= argument register count of traced function. 5185 * 5186 * long bpf_get_func_ret(void *ctx, u64 *value) 5187 * Description 5188 * Get return value of the traced function (for tracing programs) 5189 * in **value**. 5190 * 5191 * Return 5192 * 0 on success. 5193 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5194 * 5195 * long bpf_get_func_arg_cnt(void *ctx) 5196 * Description 5197 * Get number of registers of the traced function (for tracing programs) where 5198 * function arguments are stored in these registers. 5199 * 5200 * Return 5201 * The number of argument registers of the traced function. 5202 * 5203 * int bpf_get_retval(void) 5204 * Description 5205 * Get the BPF program's return value that will be returned to the upper layers. 5206 * 5207 * This helper is currently supported by cgroup programs and only by the hooks 5208 * where BPF program's return value is returned to the userspace via errno. 5209 * Return 5210 * The BPF program's return value. 5211 * 5212 * int bpf_set_retval(int retval) 5213 * Description 5214 * Set the BPF program's return value that will be returned to the upper layers. 5215 * 5216 * This helper is currently supported by cgroup programs and only by the hooks 5217 * where BPF program's return value is returned to the userspace via errno. 5218 * 5219 * Note that there is the following corner case where the program exports an error 5220 * via bpf_set_retval but signals success via 'return 1': 5221 * 5222 * bpf_set_retval(-EPERM); 5223 * return 1; 5224 * 5225 * In this case, the BPF program's return value will use helper's -EPERM. This 5226 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5227 * 5228 * Return 5229 * 0 on success, or a negative error in case of failure. 5230 * 5231 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5232 * Description 5233 * Get the total size of a given xdp buff (linear and paged area) 5234 * Return 5235 * The total size of a given xdp buffer. 5236 * 5237 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5238 * Description 5239 * This helper is provided as an easy way to load data from a 5240 * xdp buffer. It can be used to load *len* bytes from *offset* from 5241 * the frame associated to *xdp_md*, into the buffer pointed by 5242 * *buf*. 5243 * Return 5244 * 0 on success, or a negative error in case of failure. 5245 * 5246 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5247 * Description 5248 * Store *len* bytes from buffer *buf* into the frame 5249 * associated to *xdp_md*, at *offset*. 5250 * Return 5251 * 0 on success, or a negative error in case of failure. 5252 * 5253 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5254 * Description 5255 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5256 * address space, and stores the data in *dst*. *flags* is not 5257 * used yet and is provided for future extensibility. This helper 5258 * can only be used by sleepable programs. 5259 * Return 5260 * 0 on success, or a negative error in case of failure. On error 5261 * *dst* buffer is zeroed out. 5262 * 5263 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5264 * Description 5265 * Change the __sk_buff->tstamp_type to *tstamp_type* 5266 * and set *tstamp* to the __sk_buff->tstamp together. 5267 * 5268 * If there is no need to change the __sk_buff->tstamp_type, 5269 * the tstamp value can be directly written to __sk_buff->tstamp 5270 * instead. 5271 * 5272 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5273 * will be kept during bpf_redirect_*(). A non zero 5274 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5275 * *tstamp_type*. 5276 * 5277 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5278 * with a zero *tstamp*. 5279 * 5280 * Only IPv4 and IPv6 skb->protocol are supported. 5281 * 5282 * This function is most useful when it needs to set a 5283 * mono delivery time to __sk_buff->tstamp and then 5284 * bpf_redirect_*() to the egress of an iface. For example, 5285 * changing the (rcv) timestamp in __sk_buff->tstamp at 5286 * ingress to a mono delivery time and then bpf_redirect_*() 5287 * to sch_fq@phy-dev. 5288 * Return 5289 * 0 on success. 5290 * **-EINVAL** for invalid input 5291 * **-EOPNOTSUPP** for unsupported protocol 5292 * 5293 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5294 * Description 5295 * Returns a calculated IMA hash of the *file*. 5296 * If the hash is larger than *size*, then only *size* 5297 * bytes will be copied to *dst* 5298 * Return 5299 * The **hash_algo** is returned on success, 5300 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5301 * invalid arguments are passed. 5302 * 5303 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5304 * Description 5305 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5306 * old value. *ptr* can be NULL, otherwise it must be a referenced 5307 * pointer which will be released when this helper is called. 5308 * Return 5309 * The old value of kptr (which can be NULL). The returned pointer 5310 * if not NULL, is a reference which must be released using its 5311 * corresponding release function, or moved into a BPF map before 5312 * program exit. 5313 * 5314 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5315 * Description 5316 * Perform a lookup in *percpu map* for an entry associated to 5317 * *key* on *cpu*. 5318 * Return 5319 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5320 * was found or *cpu* is invalid. 5321 * 5322 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5323 * Description 5324 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5325 * Return 5326 * *sk* if casting is valid, or **NULL** otherwise. 5327 * 5328 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5329 * Description 5330 * Get a dynptr to local memory *data*. 5331 * 5332 * *data* must be a ptr to a map value. 5333 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5334 * *flags* is currently unused. 5335 * Return 5336 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5337 * -EINVAL if flags is not 0. 5338 * 5339 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5340 * Description 5341 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5342 * through the dynptr interface. *flags* must be 0. 5343 * 5344 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5345 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5346 * reservation fails. This is enforced by the verifier. 5347 * Return 5348 * 0 on success, or a negative error in case of failure. 5349 * 5350 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5351 * Description 5352 * Submit reserved ring buffer sample, pointed to by *data*, 5353 * through the dynptr interface. This is a no-op if the dynptr is 5354 * invalid/null. 5355 * 5356 * For more information on *flags*, please see 5357 * 'bpf_ringbuf_submit'. 5358 * Return 5359 * Nothing. Always succeeds. 5360 * 5361 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5362 * Description 5363 * Discard reserved ring buffer sample through the dynptr 5364 * interface. This is a no-op if the dynptr is invalid/null. 5365 * 5366 * For more information on *flags*, please see 5367 * 'bpf_ringbuf_discard'. 5368 * Return 5369 * Nothing. Always succeeds. 5370 * 5371 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5372 * Description 5373 * Read *len* bytes from *src* into *dst*, starting from *offset* 5374 * into *src*. 5375 * *flags* is currently unused. 5376 * Return 5377 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5378 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5379 * *flags* is not 0. 5380 * 5381 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5382 * Description 5383 * Write *len* bytes from *src* into *dst*, starting from *offset* 5384 * into *dst*. 5385 * 5386 * *flags* must be 0 except for skb-type dynptrs. 5387 * 5388 * For skb-type dynptrs: 5389 * * All data slices of the dynptr are automatically 5390 * invalidated after **bpf_dynptr_write**\ (). This is 5391 * because writing may pull the skb and change the 5392 * underlying packet buffer. 5393 * 5394 * * For *flags*, please see the flags accepted by 5395 * **bpf_skb_store_bytes**\ (). 5396 * Return 5397 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5398 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5399 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5400 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5401 * 5402 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5403 * Description 5404 * Get a pointer to the underlying dynptr data. 5405 * 5406 * *len* must be a statically known value. The returned data slice 5407 * is invalidated whenever the dynptr is invalidated. 5408 * 5409 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5410 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5411 * Return 5412 * Pointer to the underlying dynptr data, NULL if the dynptr is 5413 * read-only, if the dynptr is invalid, or if the offset and length 5414 * is out of bounds. 5415 * 5416 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5417 * Description 5418 * Try to issue a SYN cookie for the packet with corresponding 5419 * IPv4/TCP headers, *iph* and *th*, without depending on a 5420 * listening socket. 5421 * 5422 * *iph* points to the IPv4 header. 5423 * 5424 * *th* points to the start of the TCP header, while *th_len* 5425 * contains the length of the TCP header (at least 5426 * **sizeof**\ (**struct tcphdr**)). 5427 * Return 5428 * On success, lower 32 bits hold the generated SYN cookie in 5429 * followed by 16 bits which hold the MSS value for that cookie, 5430 * and the top 16 bits are unused. 5431 * 5432 * On failure, the returned value is one of the following: 5433 * 5434 * **-EINVAL** if *th_len* is invalid. 5435 * 5436 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5437 * Description 5438 * Try to issue a SYN cookie for the packet with corresponding 5439 * IPv6/TCP headers, *iph* and *th*, without depending on a 5440 * listening socket. 5441 * 5442 * *iph* points to the IPv6 header. 5443 * 5444 * *th* points to the start of the TCP header, while *th_len* 5445 * contains the length of the TCP header (at least 5446 * **sizeof**\ (**struct tcphdr**)). 5447 * Return 5448 * On success, lower 32 bits hold the generated SYN cookie in 5449 * followed by 16 bits which hold the MSS value for that cookie, 5450 * and the top 16 bits are unused. 5451 * 5452 * On failure, the returned value is one of the following: 5453 * 5454 * **-EINVAL** if *th_len* is invalid. 5455 * 5456 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5457 * 5458 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5459 * Description 5460 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5461 * without depending on a listening socket. 5462 * 5463 * *iph* points to the IPv4 header. 5464 * 5465 * *th* points to the TCP header. 5466 * Return 5467 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5468 * 5469 * On failure, the returned value is one of the following: 5470 * 5471 * **-EACCES** if the SYN cookie is not valid. 5472 * 5473 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5474 * Description 5475 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5476 * without depending on a listening socket. 5477 * 5478 * *iph* points to the IPv6 header. 5479 * 5480 * *th* points to the TCP header. 5481 * Return 5482 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5483 * 5484 * On failure, the returned value is one of the following: 5485 * 5486 * **-EACCES** if the SYN cookie is not valid. 5487 * 5488 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5489 * 5490 * u64 bpf_ktime_get_tai_ns(void) 5491 * Description 5492 * A nonsettable system-wide clock derived from wall-clock time but 5493 * ignoring leap seconds. This clock does not experience 5494 * discontinuities and backwards jumps caused by NTP inserting leap 5495 * seconds as CLOCK_REALTIME does. 5496 * 5497 * See: **clock_gettime**\ (**CLOCK_TAI**) 5498 * Return 5499 * Current *ktime*. 5500 * 5501 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5502 * Description 5503 * Drain samples from the specified user ring buffer, and invoke 5504 * the provided callback for each such sample: 5505 * 5506 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5507 * 5508 * If **callback_fn** returns 0, the helper will continue to try 5509 * and drain the next sample, up to a maximum of 5510 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5511 * the helper will skip the rest of the samples and return. Other 5512 * return values are not used now, and will be rejected by the 5513 * verifier. 5514 * Return 5515 * The number of drained samples if no error was encountered while 5516 * draining samples, or 0 if no samples were present in the ring 5517 * buffer. If a user-space producer was epoll-waiting on this map, 5518 * and at least one sample was drained, they will receive an event 5519 * notification notifying them of available space in the ring 5520 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5521 * function, no wakeup notification will be sent. If the 5522 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5523 * be sent even if no sample was drained. 5524 * 5525 * On failure, the returned value is one of the following: 5526 * 5527 * **-EBUSY** if the ring buffer is contended, and another calling 5528 * context was concurrently draining the ring buffer. 5529 * 5530 * **-EINVAL** if user-space is not properly tracking the ring 5531 * buffer due to the producer position not being aligned to 8 5532 * bytes, a sample not being aligned to 8 bytes, or the producer 5533 * position not matching the advertised length of a sample. 5534 * 5535 * **-E2BIG** if user-space has tried to publish a sample which is 5536 * larger than the size of the ring buffer, or which cannot fit 5537 * within a struct bpf_dynptr. 5538 * 5539 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5540 * Description 5541 * Get a bpf_local_storage from the *cgroup*. 5542 * 5543 * Logically, it could be thought of as getting the value from 5544 * a *map* with *cgroup* as the **key**. From this 5545 * perspective, the usage is not much different from 5546 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5547 * helper enforces the key must be a cgroup struct and the map must also 5548 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5549 * 5550 * In reality, the local-storage value is embedded directly inside of the 5551 * *cgroup* object itself, rather than being located in the 5552 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5553 * queried for some *map* on a *cgroup* object, the kernel will perform an 5554 * O(n) iteration over all of the live local-storage values for that 5555 * *cgroup* object until the local-storage value for the *map* is found. 5556 * 5557 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5558 * used such that a new bpf_local_storage will be 5559 * created if one does not exist. *value* can be used 5560 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5561 * the initial value of a bpf_local_storage. If *value* is 5562 * **NULL**, the new bpf_local_storage will be zero initialized. 5563 * Return 5564 * A bpf_local_storage pointer is returned on success. 5565 * 5566 * **NULL** if not found or there was an error in adding 5567 * a new bpf_local_storage. 5568 * 5569 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5570 * Description 5571 * Delete a bpf_local_storage from a *cgroup*. 5572 * Return 5573 * 0 on success. 5574 * 5575 * **-ENOENT** if the bpf_local_storage cannot be found. 5576 */ 5577#define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5578 FN(unspec, 0, ##ctx) \ 5579 FN(map_lookup_elem, 1, ##ctx) \ 5580 FN(map_update_elem, 2, ##ctx) \ 5581 FN(map_delete_elem, 3, ##ctx) \ 5582 FN(probe_read, 4, ##ctx) \ 5583 FN(ktime_get_ns, 5, ##ctx) \ 5584 FN(trace_printk, 6, ##ctx) \ 5585 FN(get_prandom_u32, 7, ##ctx) \ 5586 FN(get_smp_processor_id, 8, ##ctx) \ 5587 FN(skb_store_bytes, 9, ##ctx) \ 5588 FN(l3_csum_replace, 10, ##ctx) \ 5589 FN(l4_csum_replace, 11, ##ctx) \ 5590 FN(tail_call, 12, ##ctx) \ 5591 FN(clone_redirect, 13, ##ctx) \ 5592 FN(get_current_pid_tgid, 14, ##ctx) \ 5593 FN(get_current_uid_gid, 15, ##ctx) \ 5594 FN(get_current_comm, 16, ##ctx) \ 5595 FN(get_cgroup_classid, 17, ##ctx) \ 5596 FN(skb_vlan_push, 18, ##ctx) \ 5597 FN(skb_vlan_pop, 19, ##ctx) \ 5598 FN(skb_get_tunnel_key, 20, ##ctx) \ 5599 FN(skb_set_tunnel_key, 21, ##ctx) \ 5600 FN(perf_event_read, 22, ##ctx) \ 5601 FN(redirect, 23, ##ctx) \ 5602 FN(get_route_realm, 24, ##ctx) \ 5603 FN(perf_event_output, 25, ##ctx) \ 5604 FN(skb_load_bytes, 26, ##ctx) \ 5605 FN(get_stackid, 27, ##ctx) \ 5606 FN(csum_diff, 28, ##ctx) \ 5607 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5608 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5609 FN(skb_change_proto, 31, ##ctx) \ 5610 FN(skb_change_type, 32, ##ctx) \ 5611 FN(skb_under_cgroup, 33, ##ctx) \ 5612 FN(get_hash_recalc, 34, ##ctx) \ 5613 FN(get_current_task, 35, ##ctx) \ 5614 FN(probe_write_user, 36, ##ctx) \ 5615 FN(current_task_under_cgroup, 37, ##ctx) \ 5616 FN(skb_change_tail, 38, ##ctx) \ 5617 FN(skb_pull_data, 39, ##ctx) \ 5618 FN(csum_update, 40, ##ctx) \ 5619 FN(set_hash_invalid, 41, ##ctx) \ 5620 FN(get_numa_node_id, 42, ##ctx) \ 5621 FN(skb_change_head, 43, ##ctx) \ 5622 FN(xdp_adjust_head, 44, ##ctx) \ 5623 FN(probe_read_str, 45, ##ctx) \ 5624 FN(get_socket_cookie, 46, ##ctx) \ 5625 FN(get_socket_uid, 47, ##ctx) \ 5626 FN(set_hash, 48, ##ctx) \ 5627 FN(setsockopt, 49, ##ctx) \ 5628 FN(skb_adjust_room, 50, ##ctx) \ 5629 FN(redirect_map, 51, ##ctx) \ 5630 FN(sk_redirect_map, 52, ##ctx) \ 5631 FN(sock_map_update, 53, ##ctx) \ 5632 FN(xdp_adjust_meta, 54, ##ctx) \ 5633 FN(perf_event_read_value, 55, ##ctx) \ 5634 FN(perf_prog_read_value, 56, ##ctx) \ 5635 FN(getsockopt, 57, ##ctx) \ 5636 FN(override_return, 58, ##ctx) \ 5637 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5638 FN(msg_redirect_map, 60, ##ctx) \ 5639 FN(msg_apply_bytes, 61, ##ctx) \ 5640 FN(msg_cork_bytes, 62, ##ctx) \ 5641 FN(msg_pull_data, 63, ##ctx) \ 5642 FN(bind, 64, ##ctx) \ 5643 FN(xdp_adjust_tail, 65, ##ctx) \ 5644 FN(skb_get_xfrm_state, 66, ##ctx) \ 5645 FN(get_stack, 67, ##ctx) \ 5646 FN(skb_load_bytes_relative, 68, ##ctx) \ 5647 FN(fib_lookup, 69, ##ctx) \ 5648 FN(sock_hash_update, 70, ##ctx) \ 5649 FN(msg_redirect_hash, 71, ##ctx) \ 5650 FN(sk_redirect_hash, 72, ##ctx) \ 5651 FN(lwt_push_encap, 73, ##ctx) \ 5652 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5653 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5654 FN(lwt_seg6_action, 76, ##ctx) \ 5655 FN(rc_repeat, 77, ##ctx) \ 5656 FN(rc_keydown, 78, ##ctx) \ 5657 FN(skb_cgroup_id, 79, ##ctx) \ 5658 FN(get_current_cgroup_id, 80, ##ctx) \ 5659 FN(get_local_storage, 81, ##ctx) \ 5660 FN(sk_select_reuseport, 82, ##ctx) \ 5661 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5662 FN(sk_lookup_tcp, 84, ##ctx) \ 5663 FN(sk_lookup_udp, 85, ##ctx) \ 5664 FN(sk_release, 86, ##ctx) \ 5665 FN(map_push_elem, 87, ##ctx) \ 5666 FN(map_pop_elem, 88, ##ctx) \ 5667 FN(map_peek_elem, 89, ##ctx) \ 5668 FN(msg_push_data, 90, ##ctx) \ 5669 FN(msg_pop_data, 91, ##ctx) \ 5670 FN(rc_pointer_rel, 92, ##ctx) \ 5671 FN(spin_lock, 93, ##ctx) \ 5672 FN(spin_unlock, 94, ##ctx) \ 5673 FN(sk_fullsock, 95, ##ctx) \ 5674 FN(tcp_sock, 96, ##ctx) \ 5675 FN(skb_ecn_set_ce, 97, ##ctx) \ 5676 FN(get_listener_sock, 98, ##ctx) \ 5677 FN(skc_lookup_tcp, 99, ##ctx) \ 5678 FN(tcp_check_syncookie, 100, ##ctx) \ 5679 FN(sysctl_get_name, 101, ##ctx) \ 5680 FN(sysctl_get_current_value, 102, ##ctx) \ 5681 FN(sysctl_get_new_value, 103, ##ctx) \ 5682 FN(sysctl_set_new_value, 104, ##ctx) \ 5683 FN(strtol, 105, ##ctx) \ 5684 FN(strtoul, 106, ##ctx) \ 5685 FN(sk_storage_get, 107, ##ctx) \ 5686 FN(sk_storage_delete, 108, ##ctx) \ 5687 FN(send_signal, 109, ##ctx) \ 5688 FN(tcp_gen_syncookie, 110, ##ctx) \ 5689 FN(skb_output, 111, ##ctx) \ 5690 FN(probe_read_user, 112, ##ctx) \ 5691 FN(probe_read_kernel, 113, ##ctx) \ 5692 FN(probe_read_user_str, 114, ##ctx) \ 5693 FN(probe_read_kernel_str, 115, ##ctx) \ 5694 FN(tcp_send_ack, 116, ##ctx) \ 5695 FN(send_signal_thread, 117, ##ctx) \ 5696 FN(jiffies64, 118, ##ctx) \ 5697 FN(read_branch_records, 119, ##ctx) \ 5698 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5699 FN(xdp_output, 121, ##ctx) \ 5700 FN(get_netns_cookie, 122, ##ctx) \ 5701 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5702 FN(sk_assign, 124, ##ctx) \ 5703 FN(ktime_get_boot_ns, 125, ##ctx) \ 5704 FN(seq_printf, 126, ##ctx) \ 5705 FN(seq_write, 127, ##ctx) \ 5706 FN(sk_cgroup_id, 128, ##ctx) \ 5707 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5708 FN(ringbuf_output, 130, ##ctx) \ 5709 FN(ringbuf_reserve, 131, ##ctx) \ 5710 FN(ringbuf_submit, 132, ##ctx) \ 5711 FN(ringbuf_discard, 133, ##ctx) \ 5712 FN(ringbuf_query, 134, ##ctx) \ 5713 FN(csum_level, 135, ##ctx) \ 5714 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5715 FN(skc_to_tcp_sock, 137, ##ctx) \ 5716 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5717 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5718 FN(skc_to_udp6_sock, 140, ##ctx) \ 5719 FN(get_task_stack, 141, ##ctx) \ 5720 FN(load_hdr_opt, 142, ##ctx) \ 5721 FN(store_hdr_opt, 143, ##ctx) \ 5722 FN(reserve_hdr_opt, 144, ##ctx) \ 5723 FN(inode_storage_get, 145, ##ctx) \ 5724 FN(inode_storage_delete, 146, ##ctx) \ 5725 FN(d_path, 147, ##ctx) \ 5726 FN(copy_from_user, 148, ##ctx) \ 5727 FN(snprintf_btf, 149, ##ctx) \ 5728 FN(seq_printf_btf, 150, ##ctx) \ 5729 FN(skb_cgroup_classid, 151, ##ctx) \ 5730 FN(redirect_neigh, 152, ##ctx) \ 5731 FN(per_cpu_ptr, 153, ##ctx) \ 5732 FN(this_cpu_ptr, 154, ##ctx) \ 5733 FN(redirect_peer, 155, ##ctx) \ 5734 FN(task_storage_get, 156, ##ctx) \ 5735 FN(task_storage_delete, 157, ##ctx) \ 5736 FN(get_current_task_btf, 158, ##ctx) \ 5737 FN(bprm_opts_set, 159, ##ctx) \ 5738 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5739 FN(ima_inode_hash, 161, ##ctx) \ 5740 FN(sock_from_file, 162, ##ctx) \ 5741 FN(check_mtu, 163, ##ctx) \ 5742 FN(for_each_map_elem, 164, ##ctx) \ 5743 FN(snprintf, 165, ##ctx) \ 5744 FN(sys_bpf, 166, ##ctx) \ 5745 FN(btf_find_by_name_kind, 167, ##ctx) \ 5746 FN(sys_close, 168, ##ctx) \ 5747 FN(timer_init, 169, ##ctx) \ 5748 FN(timer_set_callback, 170, ##ctx) \ 5749 FN(timer_start, 171, ##ctx) \ 5750 FN(timer_cancel, 172, ##ctx) \ 5751 FN(get_func_ip, 173, ##ctx) \ 5752 FN(get_attach_cookie, 174, ##ctx) \ 5753 FN(task_pt_regs, 175, ##ctx) \ 5754 FN(get_branch_snapshot, 176, ##ctx) \ 5755 FN(trace_vprintk, 177, ##ctx) \ 5756 FN(skc_to_unix_sock, 178, ##ctx) \ 5757 FN(kallsyms_lookup_name, 179, ##ctx) \ 5758 FN(find_vma, 180, ##ctx) \ 5759 FN(loop, 181, ##ctx) \ 5760 FN(strncmp, 182, ##ctx) \ 5761 FN(get_func_arg, 183, ##ctx) \ 5762 FN(get_func_ret, 184, ##ctx) \ 5763 FN(get_func_arg_cnt, 185, ##ctx) \ 5764 FN(get_retval, 186, ##ctx) \ 5765 FN(set_retval, 187, ##ctx) \ 5766 FN(xdp_get_buff_len, 188, ##ctx) \ 5767 FN(xdp_load_bytes, 189, ##ctx) \ 5768 FN(xdp_store_bytes, 190, ##ctx) \ 5769 FN(copy_from_user_task, 191, ##ctx) \ 5770 FN(skb_set_tstamp, 192, ##ctx) \ 5771 FN(ima_file_hash, 193, ##ctx) \ 5772 FN(kptr_xchg, 194, ##ctx) \ 5773 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5774 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5775 FN(dynptr_from_mem, 197, ##ctx) \ 5776 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5777 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5778 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5779 FN(dynptr_read, 201, ##ctx) \ 5780 FN(dynptr_write, 202, ##ctx) \ 5781 FN(dynptr_data, 203, ##ctx) \ 5782 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5783 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5784 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5785 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5786 FN(ktime_get_tai_ns, 208, ##ctx) \ 5787 FN(user_ringbuf_drain, 209, ##ctx) \ 5788 FN(cgrp_storage_get, 210, ##ctx) \ 5789 FN(cgrp_storage_delete, 211, ##ctx) \ 5790 /* */ 5791 5792/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 5793 * know or care about integer value that is now passed as second argument 5794 */ 5795#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 5796#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 5797 5798/* integer value in 'imm' field of BPF_CALL instruction selects which helper 5799 * function eBPF program intends to call 5800 */ 5801#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5802enum bpf_func_id { 5803 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5804 __BPF_FUNC_MAX_ID, 5805}; 5806#undef __BPF_ENUM_FN 5807 5808/* All flags used by eBPF helper functions, placed here. */ 5809 5810/* BPF_FUNC_skb_store_bytes flags. */ 5811enum { 5812 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5813 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5814}; 5815 5816/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5817 * First 4 bits are for passing the header field size. 5818 */ 5819enum { 5820 BPF_F_HDR_FIELD_MASK = 0xfULL, 5821}; 5822 5823/* BPF_FUNC_l4_csum_replace flags. */ 5824enum { 5825 BPF_F_PSEUDO_HDR = (1ULL << 4), 5826 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5827 BPF_F_MARK_ENFORCE = (1ULL << 6), 5828}; 5829 5830/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5831enum { 5832 BPF_F_INGRESS = (1ULL << 0), 5833}; 5834 5835/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5836enum { 5837 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5838}; 5839 5840/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5841enum { 5842 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5843 BPF_F_USER_STACK = (1ULL << 8), 5844/* flags used by BPF_FUNC_get_stackid only. */ 5845 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5846 BPF_F_REUSE_STACKID = (1ULL << 10), 5847/* flags used by BPF_FUNC_get_stack only. */ 5848 BPF_F_USER_BUILD_ID = (1ULL << 11), 5849}; 5850 5851/* BPF_FUNC_skb_set_tunnel_key flags. */ 5852enum { 5853 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5854 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5855 BPF_F_SEQ_NUMBER = (1ULL << 3), 5856 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 5857}; 5858 5859/* BPF_FUNC_skb_get_tunnel_key flags. */ 5860enum { 5861 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 5862}; 5863 5864/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5865 * BPF_FUNC_perf_event_read_value flags. 5866 */ 5867enum { 5868 BPF_F_INDEX_MASK = 0xffffffffULL, 5869 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5870/* BPF_FUNC_perf_event_output for sk_buff input context. */ 5871 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5872}; 5873 5874/* Current network namespace */ 5875enum { 5876 BPF_F_CURRENT_NETNS = (-1L), 5877}; 5878 5879/* BPF_FUNC_csum_level level values. */ 5880enum { 5881 BPF_CSUM_LEVEL_QUERY, 5882 BPF_CSUM_LEVEL_INC, 5883 BPF_CSUM_LEVEL_DEC, 5884 BPF_CSUM_LEVEL_RESET, 5885}; 5886 5887/* BPF_FUNC_skb_adjust_room flags. */ 5888enum { 5889 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5890 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5891 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5892 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5893 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5894 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5895 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5896 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 5897 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 5898}; 5899 5900enum { 5901 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5902 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5903}; 5904 5905#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5906 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5907 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5908 5909/* BPF_FUNC_sysctl_get_name flags. */ 5910enum { 5911 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5912}; 5913 5914/* BPF_FUNC_<kernel_obj>_storage_get flags */ 5915enum { 5916 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5917 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5918 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5919 */ 5920 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5921}; 5922 5923/* BPF_FUNC_read_branch_records flags. */ 5924enum { 5925 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5926}; 5927 5928/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5929 * BPF_FUNC_bpf_ringbuf_output flags. 5930 */ 5931enum { 5932 BPF_RB_NO_WAKEUP = (1ULL << 0), 5933 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5934}; 5935 5936/* BPF_FUNC_bpf_ringbuf_query flags */ 5937enum { 5938 BPF_RB_AVAIL_DATA = 0, 5939 BPF_RB_RING_SIZE = 1, 5940 BPF_RB_CONS_POS = 2, 5941 BPF_RB_PROD_POS = 3, 5942}; 5943 5944/* BPF ring buffer constants */ 5945enum { 5946 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5947 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5948 BPF_RINGBUF_HDR_SZ = 8, 5949}; 5950 5951/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5952enum { 5953 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5954 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5955}; 5956 5957/* Mode for BPF_FUNC_skb_adjust_room helper. */ 5958enum bpf_adj_room_mode { 5959 BPF_ADJ_ROOM_NET, 5960 BPF_ADJ_ROOM_MAC, 5961}; 5962 5963/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5964enum bpf_hdr_start_off { 5965 BPF_HDR_START_MAC, 5966 BPF_HDR_START_NET, 5967}; 5968 5969/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5970enum bpf_lwt_encap_mode { 5971 BPF_LWT_ENCAP_SEG6, 5972 BPF_LWT_ENCAP_SEG6_INLINE, 5973 BPF_LWT_ENCAP_IP, 5974}; 5975 5976/* Flags for bpf_bprm_opts_set helper */ 5977enum { 5978 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5979}; 5980 5981/* Flags for bpf_redirect_map helper */ 5982enum { 5983 BPF_F_BROADCAST = (1ULL << 3), 5984 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5985}; 5986 5987#define __bpf_md_ptr(type, name) \ 5988union { \ 5989 type name; \ 5990 __u64 :64; \ 5991} __attribute__((aligned(8))) 5992 5993enum { 5994 BPF_SKB_TSTAMP_UNSPEC, 5995 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 5996 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 5997 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 5998 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 5999 */ 6000}; 6001 6002/* user accessible mirror of in-kernel sk_buff. 6003 * new fields can only be added to the end of this structure 6004 */ 6005struct __sk_buff { 6006 __u32 len; 6007 __u32 pkt_type; 6008 __u32 mark; 6009 __u32 queue_mapping; 6010 __u32 protocol; 6011 __u32 vlan_present; 6012 __u32 vlan_tci; 6013 __u32 vlan_proto; 6014 __u32 priority; 6015 __u32 ingress_ifindex; 6016 __u32 ifindex; 6017 __u32 tc_index; 6018 __u32 cb[5]; 6019 __u32 hash; 6020 __u32 tc_classid; 6021 __u32 data; 6022 __u32 data_end; 6023 __u32 napi_id; 6024 6025 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6026 __u32 family; 6027 __u32 remote_ip4; /* Stored in network byte order */ 6028 __u32 local_ip4; /* Stored in network byte order */ 6029 __u32 remote_ip6[4]; /* Stored in network byte order */ 6030 __u32 local_ip6[4]; /* Stored in network byte order */ 6031 __u32 remote_port; /* Stored in network byte order */ 6032 __u32 local_port; /* stored in host byte order */ 6033 /* ... here. */ 6034 6035 __u32 data_meta; 6036 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6037 __u64 tstamp; 6038 __u32 wire_len; 6039 __u32 gso_segs; 6040 __bpf_md_ptr(struct bpf_sock *, sk); 6041 __u32 gso_size; 6042 __u8 tstamp_type; 6043 __u32 :24; /* Padding, future use. */ 6044 __u64 hwtstamp; 6045}; 6046 6047struct bpf_tunnel_key { 6048 __u32 tunnel_id; 6049 union { 6050 __u32 remote_ipv4; 6051 __u32 remote_ipv6[4]; 6052 }; 6053 __u8 tunnel_tos; 6054 __u8 tunnel_ttl; 6055 union { 6056 __u16 tunnel_ext; /* compat */ 6057 __be16 tunnel_flags; 6058 }; 6059 __u32 tunnel_label; 6060 union { 6061 __u32 local_ipv4; 6062 __u32 local_ipv6[4]; 6063 }; 6064}; 6065 6066/* user accessible mirror of in-kernel xfrm_state. 6067 * new fields can only be added to the end of this structure 6068 */ 6069struct bpf_xfrm_state { 6070 __u32 reqid; 6071 __u32 spi; /* Stored in network byte order */ 6072 __u16 family; 6073 __u16 ext; /* Padding, future use. */ 6074 union { 6075 __u32 remote_ipv4; /* Stored in network byte order */ 6076 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6077 }; 6078}; 6079 6080/* Generic BPF return codes which all BPF program types may support. 6081 * The values are binary compatible with their TC_ACT_* counter-part to 6082 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6083 * programs. 6084 * 6085 * XDP is handled seprately, see XDP_*. 6086 */ 6087enum bpf_ret_code { 6088 BPF_OK = 0, 6089 /* 1 reserved */ 6090 BPF_DROP = 2, 6091 /* 3-6 reserved */ 6092 BPF_REDIRECT = 7, 6093 /* >127 are reserved for prog type specific return codes. 6094 * 6095 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6096 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6097 * changed and should be routed based on its new L3 header. 6098 * (This is an L3 redirect, as opposed to L2 redirect 6099 * represented by BPF_REDIRECT above). 6100 */ 6101 BPF_LWT_REROUTE = 128, 6102 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6103 * to indicate that no custom dissection was performed, and 6104 * fallback to standard dissector is requested. 6105 */ 6106 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6107}; 6108 6109struct bpf_sock { 6110 __u32 bound_dev_if; 6111 __u32 family; 6112 __u32 type; 6113 __u32 protocol; 6114 __u32 mark; 6115 __u32 priority; 6116 /* IP address also allows 1 and 2 bytes access */ 6117 __u32 src_ip4; 6118 __u32 src_ip6[4]; 6119 __u32 src_port; /* host byte order */ 6120 __be16 dst_port; /* network byte order */ 6121 __u16 :16; /* zero padding */ 6122 __u32 dst_ip4; 6123 __u32 dst_ip6[4]; 6124 __u32 state; 6125 __s32 rx_queue_mapping; 6126}; 6127 6128struct bpf_tcp_sock { 6129 __u32 snd_cwnd; /* Sending congestion window */ 6130 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6131 __u32 rtt_min; 6132 __u32 snd_ssthresh; /* Slow start size threshold */ 6133 __u32 rcv_nxt; /* What we want to receive next */ 6134 __u32 snd_nxt; /* Next sequence we send */ 6135 __u32 snd_una; /* First byte we want an ack for */ 6136 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6137 __u32 ecn_flags; /* ECN status bits. */ 6138 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6139 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6140 __u32 packets_out; /* Packets which are "in flight" */ 6141 __u32 retrans_out; /* Retransmitted packets out */ 6142 __u32 total_retrans; /* Total retransmits for entire connection */ 6143 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6144 * total number of segments in. 6145 */ 6146 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6147 * total number of data segments in. 6148 */ 6149 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6150 * The total number of segments sent. 6151 */ 6152 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6153 * total number of data segments sent. 6154 */ 6155 __u32 lost_out; /* Lost packets */ 6156 __u32 sacked_out; /* SACK'd packets */ 6157 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6158 * sum(delta(rcv_nxt)), or how many bytes 6159 * were acked. 6160 */ 6161 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6162 * sum(delta(snd_una)), or how many bytes 6163 * were acked. 6164 */ 6165 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6166 * total number of DSACK blocks received 6167 */ 6168 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6169 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6170 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6171}; 6172 6173struct bpf_sock_tuple { 6174 union { 6175 struct { 6176 __be32 saddr; 6177 __be32 daddr; 6178 __be16 sport; 6179 __be16 dport; 6180 } ipv4; 6181 struct { 6182 __be32 saddr[4]; 6183 __be32 daddr[4]; 6184 __be16 sport; 6185 __be16 dport; 6186 } ipv6; 6187 }; 6188}; 6189 6190struct bpf_xdp_sock { 6191 __u32 queue_id; 6192}; 6193 6194#define XDP_PACKET_HEADROOM 256 6195 6196/* User return codes for XDP prog type. 6197 * A valid XDP program must return one of these defined values. All other 6198 * return codes are reserved for future use. Unknown return codes will 6199 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6200 */ 6201enum xdp_action { 6202 XDP_ABORTED = 0, 6203 XDP_DROP, 6204 XDP_PASS, 6205 XDP_TX, 6206 XDP_REDIRECT, 6207}; 6208 6209/* user accessible metadata for XDP packet hook 6210 * new fields must be added to the end of this structure 6211 */ 6212struct xdp_md { 6213 __u32 data; 6214 __u32 data_end; 6215 __u32 data_meta; 6216 /* Below access go through struct xdp_rxq_info */ 6217 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6218 __u32 rx_queue_index; /* rxq->queue_index */ 6219 6220 __u32 egress_ifindex; /* txq->dev->ifindex */ 6221}; 6222 6223/* DEVMAP map-value layout 6224 * 6225 * The struct data-layout of map-value is a configuration interface. 6226 * New members can only be added to the end of this structure. 6227 */ 6228struct bpf_devmap_val { 6229 __u32 ifindex; /* device index */ 6230 union { 6231 int fd; /* prog fd on map write */ 6232 __u32 id; /* prog id on map read */ 6233 } bpf_prog; 6234}; 6235 6236/* CPUMAP map-value layout 6237 * 6238 * The struct data-layout of map-value is a configuration interface. 6239 * New members can only be added to the end of this structure. 6240 */ 6241struct bpf_cpumap_val { 6242 __u32 qsize; /* queue size to remote target CPU */ 6243 union { 6244 int fd; /* prog fd on map write */ 6245 __u32 id; /* prog id on map read */ 6246 } bpf_prog; 6247}; 6248 6249enum sk_action { 6250 SK_DROP = 0, 6251 SK_PASS, 6252}; 6253 6254/* user accessible metadata for SK_MSG packet hook, new fields must 6255 * be added to the end of this structure 6256 */ 6257struct sk_msg_md { 6258 __bpf_md_ptr(void *, data); 6259 __bpf_md_ptr(void *, data_end); 6260 6261 __u32 family; 6262 __u32 remote_ip4; /* Stored in network byte order */ 6263 __u32 local_ip4; /* Stored in network byte order */ 6264 __u32 remote_ip6[4]; /* Stored in network byte order */ 6265 __u32 local_ip6[4]; /* Stored in network byte order */ 6266 __u32 remote_port; /* Stored in network byte order */ 6267 __u32 local_port; /* stored in host byte order */ 6268 __u32 size; /* Total size of sk_msg */ 6269 6270 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6271}; 6272 6273struct sk_reuseport_md { 6274 /* 6275 * Start of directly accessible data. It begins from 6276 * the tcp/udp header. 6277 */ 6278 __bpf_md_ptr(void *, data); 6279 /* End of directly accessible data */ 6280 __bpf_md_ptr(void *, data_end); 6281 /* 6282 * Total length of packet (starting from the tcp/udp header). 6283 * Note that the directly accessible bytes (data_end - data) 6284 * could be less than this "len". Those bytes could be 6285 * indirectly read by a helper "bpf_skb_load_bytes()". 6286 */ 6287 __u32 len; 6288 /* 6289 * Eth protocol in the mac header (network byte order). e.g. 6290 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6291 */ 6292 __u32 eth_protocol; 6293 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6294 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6295 __u32 hash; /* A hash of the packet 4 tuples */ 6296 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6297 * new incoming connection request (e.g. selecting a listen sk for 6298 * the received SYN in the TCP case). reuse->sk is one of the sk 6299 * in the reuseport group. The bpf prog can use reuse->sk to learn 6300 * the local listening ip/port without looking into the skb. 6301 * 6302 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6303 * reuse->migrating_sk is the socket that needs to be migrated 6304 * to another listening socket. migrating_sk could be a fullsock 6305 * sk that is fully established or a reqsk that is in-the-middle 6306 * of 3-way handshake. 6307 */ 6308 __bpf_md_ptr(struct bpf_sock *, sk); 6309 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6310}; 6311 6312#define BPF_TAG_SIZE 8 6313 6314struct bpf_prog_info { 6315 __u32 type; 6316 __u32 id; 6317 __u8 tag[BPF_TAG_SIZE]; 6318 __u32 jited_prog_len; 6319 __u32 xlated_prog_len; 6320 __aligned_u64 jited_prog_insns; 6321 __aligned_u64 xlated_prog_insns; 6322 __u64 load_time; /* ns since boottime */ 6323 __u32 created_by_uid; 6324 __u32 nr_map_ids; 6325 __aligned_u64 map_ids; 6326 char name[BPF_OBJ_NAME_LEN]; 6327 __u32 ifindex; 6328 __u32 gpl_compatible:1; 6329 __u32 :31; /* alignment pad */ 6330 __u64 netns_dev; 6331 __u64 netns_ino; 6332 __u32 nr_jited_ksyms; 6333 __u32 nr_jited_func_lens; 6334 __aligned_u64 jited_ksyms; 6335 __aligned_u64 jited_func_lens; 6336 __u32 btf_id; 6337 __u32 func_info_rec_size; 6338 __aligned_u64 func_info; 6339 __u32 nr_func_info; 6340 __u32 nr_line_info; 6341 __aligned_u64 line_info; 6342 __aligned_u64 jited_line_info; 6343 __u32 nr_jited_line_info; 6344 __u32 line_info_rec_size; 6345 __u32 jited_line_info_rec_size; 6346 __u32 nr_prog_tags; 6347 __aligned_u64 prog_tags; 6348 __u64 run_time_ns; 6349 __u64 run_cnt; 6350 __u64 recursion_misses; 6351 __u32 verified_insns; 6352 __u32 attach_btf_obj_id; 6353 __u32 attach_btf_id; 6354} __attribute__((aligned(8))); 6355 6356struct bpf_map_info { 6357 __u32 type; 6358 __u32 id; 6359 __u32 key_size; 6360 __u32 value_size; 6361 __u32 max_entries; 6362 __u32 map_flags; 6363 char name[BPF_OBJ_NAME_LEN]; 6364 __u32 ifindex; 6365 __u32 btf_vmlinux_value_type_id; 6366 __u64 netns_dev; 6367 __u64 netns_ino; 6368 __u32 btf_id; 6369 __u32 btf_key_type_id; 6370 __u32 btf_value_type_id; 6371 __u32 :32; /* alignment pad */ 6372 __u64 map_extra; 6373} __attribute__((aligned(8))); 6374 6375struct bpf_btf_info { 6376 __aligned_u64 btf; 6377 __u32 btf_size; 6378 __u32 id; 6379 __aligned_u64 name; 6380 __u32 name_len; 6381 __u32 kernel_btf; 6382} __attribute__((aligned(8))); 6383 6384struct bpf_link_info { 6385 __u32 type; 6386 __u32 id; 6387 __u32 prog_id; 6388 union { 6389 struct { 6390 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6391 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6392 } raw_tracepoint; 6393 struct { 6394 __u32 attach_type; 6395 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6396 __u32 target_btf_id; /* BTF type id inside the object */ 6397 } tracing; 6398 struct { 6399 __u64 cgroup_id; 6400 __u32 attach_type; 6401 } cgroup; 6402 struct { 6403 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6404 __u32 target_name_len; /* in/out: target_name buffer len */ 6405 6406 /* If the iter specific field is 32 bits, it can be put 6407 * in the first or second union. Otherwise it should be 6408 * put in the second union. 6409 */ 6410 union { 6411 struct { 6412 __u32 map_id; 6413 } map; 6414 }; 6415 union { 6416 struct { 6417 __u64 cgroup_id; 6418 __u32 order; 6419 } cgroup; 6420 struct { 6421 __u32 tid; 6422 __u32 pid; 6423 } task; 6424 }; 6425 } iter; 6426 struct { 6427 __u32 netns_ino; 6428 __u32 attach_type; 6429 } netns; 6430 struct { 6431 __u32 ifindex; 6432 } xdp; 6433 struct { 6434 __u32 map_id; 6435 } struct_ops; 6436 struct { 6437 __u32 pf; 6438 __u32 hooknum; 6439 __s32 priority; 6440 __u32 flags; 6441 } netfilter; 6442 }; 6443} __attribute__((aligned(8))); 6444 6445/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6446 * by user and intended to be used by socket (e.g. to bind to, depends on 6447 * attach type). 6448 */ 6449struct bpf_sock_addr { 6450 __u32 user_family; /* Allows 4-byte read, but no write. */ 6451 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6452 * Stored in network byte order. 6453 */ 6454 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6455 * Stored in network byte order. 6456 */ 6457 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6458 * Stored in network byte order 6459 */ 6460 __u32 family; /* Allows 4-byte read, but no write */ 6461 __u32 type; /* Allows 4-byte read, but no write */ 6462 __u32 protocol; /* Allows 4-byte read, but no write */ 6463 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6464 * Stored in network byte order. 6465 */ 6466 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6467 * Stored in network byte order. 6468 */ 6469 __bpf_md_ptr(struct bpf_sock *, sk); 6470}; 6471 6472/* User bpf_sock_ops struct to access socket values and specify request ops 6473 * and their replies. 6474 * Some of this fields are in network (bigendian) byte order and may need 6475 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6476 * New fields can only be added at the end of this structure 6477 */ 6478struct bpf_sock_ops { 6479 __u32 op; 6480 union { 6481 __u32 args[4]; /* Optionally passed to bpf program */ 6482 __u32 reply; /* Returned by bpf program */ 6483 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6484 }; 6485 __u32 family; 6486 __u32 remote_ip4; /* Stored in network byte order */ 6487 __u32 local_ip4; /* Stored in network byte order */ 6488 __u32 remote_ip6[4]; /* Stored in network byte order */ 6489 __u32 local_ip6[4]; /* Stored in network byte order */ 6490 __u32 remote_port; /* Stored in network byte order */ 6491 __u32 local_port; /* stored in host byte order */ 6492 __u32 is_fullsock; /* Some TCP fields are only valid if 6493 * there is a full socket. If not, the 6494 * fields read as zero. 6495 */ 6496 __u32 snd_cwnd; 6497 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6498 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6499 __u32 state; 6500 __u32 rtt_min; 6501 __u32 snd_ssthresh; 6502 __u32 rcv_nxt; 6503 __u32 snd_nxt; 6504 __u32 snd_una; 6505 __u32 mss_cache; 6506 __u32 ecn_flags; 6507 __u32 rate_delivered; 6508 __u32 rate_interval_us; 6509 __u32 packets_out; 6510 __u32 retrans_out; 6511 __u32 total_retrans; 6512 __u32 segs_in; 6513 __u32 data_segs_in; 6514 __u32 segs_out; 6515 __u32 data_segs_out; 6516 __u32 lost_out; 6517 __u32 sacked_out; 6518 __u32 sk_txhash; 6519 __u64 bytes_received; 6520 __u64 bytes_acked; 6521 __bpf_md_ptr(struct bpf_sock *, sk); 6522 /* [skb_data, skb_data_end) covers the whole TCP header. 6523 * 6524 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6525 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6526 * header has not been written. 6527 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6528 * been written so far. 6529 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6530 * the 3WHS. 6531 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6532 * the 3WHS. 6533 * 6534 * bpf_load_hdr_opt() can also be used to read a particular option. 6535 */ 6536 __bpf_md_ptr(void *, skb_data); 6537 __bpf_md_ptr(void *, skb_data_end); 6538 __u32 skb_len; /* The total length of a packet. 6539 * It includes the header, options, 6540 * and payload. 6541 */ 6542 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6543 * an easy way to check for tcp_flags 6544 * without parsing skb_data. 6545 * 6546 * In particular, the skb_tcp_flags 6547 * will still be available in 6548 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6549 * the outgoing header has not 6550 * been written yet. 6551 */ 6552 __u64 skb_hwtstamp; 6553}; 6554 6555/* Definitions for bpf_sock_ops_cb_flags */ 6556enum { 6557 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6558 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6559 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6560 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6561 /* Call bpf for all received TCP headers. The bpf prog will be 6562 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6563 * 6564 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6565 * for the header option related helpers that will be useful 6566 * to the bpf programs. 6567 * 6568 * It could be used at the client/active side (i.e. connect() side) 6569 * when the server told it that the server was in syncookie 6570 * mode and required the active side to resend the bpf-written 6571 * options. The active side can keep writing the bpf-options until 6572 * it received a valid packet from the server side to confirm 6573 * the earlier packet (and options) has been received. The later 6574 * example patch is using it like this at the active side when the 6575 * server is in syncookie mode. 6576 * 6577 * The bpf prog will usually turn this off in the common cases. 6578 */ 6579 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6580 /* Call bpf when kernel has received a header option that 6581 * the kernel cannot handle. The bpf prog will be called under 6582 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6583 * 6584 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6585 * for the header option related helpers that will be useful 6586 * to the bpf programs. 6587 */ 6588 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6589 /* Call bpf when the kernel is writing header options for the 6590 * outgoing packet. The bpf prog will first be called 6591 * to reserve space in a skb under 6592 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6593 * the bpf prog will be called to write the header option(s) 6594 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6595 * 6596 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6597 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6598 * related helpers that will be useful to the bpf programs. 6599 * 6600 * The kernel gets its chance to reserve space and write 6601 * options first before the BPF program does. 6602 */ 6603 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6604/* Mask of all currently supported cb flags */ 6605 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6606}; 6607 6608/* List of known BPF sock_ops operators. 6609 * New entries can only be added at the end 6610 */ 6611enum { 6612 BPF_SOCK_OPS_VOID, 6613 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6614 * -1 if default value should be used 6615 */ 6616 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6617 * window (in packets) or -1 if default 6618 * value should be used 6619 */ 6620 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6621 * active connection is initialized 6622 */ 6623 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6624 * active connection is 6625 * established 6626 */ 6627 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6628 * passive connection is 6629 * established 6630 */ 6631 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6632 * needs ECN 6633 */ 6634 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6635 * based on the path and may be 6636 * dependent on the congestion control 6637 * algorithm. In general it indicates 6638 * a congestion threshold. RTTs above 6639 * this indicate congestion 6640 */ 6641 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6642 * Arg1: value of icsk_retransmits 6643 * Arg2: value of icsk_rto 6644 * Arg3: whether RTO has expired 6645 */ 6646 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6647 * Arg1: sequence number of 1st byte 6648 * Arg2: # segments 6649 * Arg3: return value of 6650 * tcp_transmit_skb (0 => success) 6651 */ 6652 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6653 * Arg1: old_state 6654 * Arg2: new_state 6655 */ 6656 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6657 * socket transition to LISTEN state. 6658 */ 6659 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6660 */ 6661 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6662 * It will be called to handle 6663 * the packets received at 6664 * an already established 6665 * connection. 6666 * 6667 * sock_ops->skb_data: 6668 * Referring to the received skb. 6669 * It covers the TCP header only. 6670 * 6671 * bpf_load_hdr_opt() can also 6672 * be used to search for a 6673 * particular option. 6674 */ 6675 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6676 * header option later in 6677 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6678 * Arg1: bool want_cookie. (in 6679 * writing SYNACK only) 6680 * 6681 * sock_ops->skb_data: 6682 * Not available because no header has 6683 * been written yet. 6684 * 6685 * sock_ops->skb_tcp_flags: 6686 * The tcp_flags of the 6687 * outgoing skb. (e.g. SYN, ACK, FIN). 6688 * 6689 * bpf_reserve_hdr_opt() should 6690 * be used to reserve space. 6691 */ 6692 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6693 * Arg1: bool want_cookie. (in 6694 * writing SYNACK only) 6695 * 6696 * sock_ops->skb_data: 6697 * Referring to the outgoing skb. 6698 * It covers the TCP header 6699 * that has already been written 6700 * by the kernel and the 6701 * earlier bpf-progs. 6702 * 6703 * sock_ops->skb_tcp_flags: 6704 * The tcp_flags of the outgoing 6705 * skb. (e.g. SYN, ACK, FIN). 6706 * 6707 * bpf_store_hdr_opt() should 6708 * be used to write the 6709 * option. 6710 * 6711 * bpf_load_hdr_opt() can also 6712 * be used to search for a 6713 * particular option that 6714 * has already been written 6715 * by the kernel or the 6716 * earlier bpf-progs. 6717 */ 6718}; 6719 6720/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6721 * changes between the TCP and BPF versions. Ideally this should never happen. 6722 * If it does, we need to add code to convert them before calling 6723 * the BPF sock_ops function. 6724 */ 6725enum { 6726 BPF_TCP_ESTABLISHED = 1, 6727 BPF_TCP_SYN_SENT, 6728 BPF_TCP_SYN_RECV, 6729 BPF_TCP_FIN_WAIT1, 6730 BPF_TCP_FIN_WAIT2, 6731 BPF_TCP_TIME_WAIT, 6732 BPF_TCP_CLOSE, 6733 BPF_TCP_CLOSE_WAIT, 6734 BPF_TCP_LAST_ACK, 6735 BPF_TCP_LISTEN, 6736 BPF_TCP_CLOSING, /* Now a valid state */ 6737 BPF_TCP_NEW_SYN_RECV, 6738 6739 BPF_TCP_MAX_STATES /* Leave at the end! */ 6740}; 6741 6742enum { 6743 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6744 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6745 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6746 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6747 /* Copy the SYN pkt to optval 6748 * 6749 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6750 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6751 * to only getting from the saved_syn. It can either get the 6752 * syn packet from: 6753 * 6754 * 1. the just-received SYN packet (only available when writing the 6755 * SYNACK). It will be useful when it is not necessary to 6756 * save the SYN packet for latter use. It is also the only way 6757 * to get the SYN during syncookie mode because the syn 6758 * packet cannot be saved during syncookie. 6759 * 6760 * OR 6761 * 6762 * 2. the earlier saved syn which was done by 6763 * bpf_setsockopt(TCP_SAVE_SYN). 6764 * 6765 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6766 * SYN packet is obtained. 6767 * 6768 * If the bpf-prog does not need the IP[46] header, the 6769 * bpf-prog can avoid parsing the IP header by using 6770 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6771 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6772 * 6773 * >0: Total number of bytes copied 6774 * -ENOSPC: Not enough space in optval. Only optlen number of 6775 * bytes is copied. 6776 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6777 * is not saved by setsockopt(TCP_SAVE_SYN). 6778 */ 6779 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6780 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6781 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6782}; 6783 6784enum { 6785 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6786}; 6787 6788/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6789 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6790 */ 6791enum { 6792 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6793 * total option spaces 6794 * required for an established 6795 * sk in order to calculate the 6796 * MSS. No skb is actually 6797 * sent. 6798 */ 6799 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6800 * when sending a SYN. 6801 */ 6802}; 6803 6804struct bpf_perf_event_value { 6805 __u64 counter; 6806 __u64 enabled; 6807 __u64 running; 6808}; 6809 6810enum { 6811 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6812 BPF_DEVCG_ACC_READ = (1ULL << 1), 6813 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 6814}; 6815 6816enum { 6817 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 6818 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 6819}; 6820 6821struct bpf_cgroup_dev_ctx { 6822 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 6823 __u32 access_type; 6824 __u32 major; 6825 __u32 minor; 6826}; 6827 6828struct bpf_raw_tracepoint_args { 6829 __u64 args[0]; 6830}; 6831 6832/* DIRECT: Skip the FIB rules and go to FIB table associated with device 6833 * OUTPUT: Do lookup from egress perspective; default is ingress 6834 */ 6835enum { 6836 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6837 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6838 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 6839 BPF_FIB_LOOKUP_TBID = (1U << 3), 6840}; 6841 6842enum { 6843 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6844 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6845 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6846 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6847 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6848 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6849 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6850 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6851 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6852}; 6853 6854struct bpf_fib_lookup { 6855 /* input: network family for lookup (AF_INET, AF_INET6) 6856 * output: network family of egress nexthop 6857 */ 6858 __u8 family; 6859 6860 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6861 __u8 l4_protocol; 6862 __be16 sport; 6863 __be16 dport; 6864 6865 union { /* used for MTU check */ 6866 /* input to lookup */ 6867 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6868 6869 /* output: MTU value */ 6870 __u16 mtu_result; 6871 }; 6872 /* input: L3 device index for lookup 6873 * output: device index from FIB lookup 6874 */ 6875 __u32 ifindex; 6876 6877 union { 6878 /* inputs to lookup */ 6879 __u8 tos; /* AF_INET */ 6880 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6881 6882 /* output: metric of fib result (IPv4/IPv6 only) */ 6883 __u32 rt_metric; 6884 }; 6885 6886 union { 6887 __be32 ipv4_src; 6888 __u32 ipv6_src[4]; /* in6_addr; network order */ 6889 }; 6890 6891 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6892 * network header. output: bpf_fib_lookup sets to gateway address 6893 * if FIB lookup returns gateway route 6894 */ 6895 union { 6896 __be32 ipv4_dst; 6897 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6898 }; 6899 6900 union { 6901 struct { 6902 /* output */ 6903 __be16 h_vlan_proto; 6904 __be16 h_vlan_TCI; 6905 }; 6906 /* input: when accompanied with the 6907 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 6908 * specific routing table to use for the fib lookup. 6909 */ 6910 __u32 tbid; 6911 }; 6912 6913 __u8 smac[6]; /* ETH_ALEN */ 6914 __u8 dmac[6]; /* ETH_ALEN */ 6915}; 6916 6917struct bpf_redir_neigh { 6918 /* network family for lookup (AF_INET, AF_INET6) */ 6919 __u32 nh_family; 6920 /* network address of nexthop; skips fib lookup to find gateway */ 6921 union { 6922 __be32 ipv4_nh; 6923 __u32 ipv6_nh[4]; /* in6_addr; network order */ 6924 }; 6925}; 6926 6927/* bpf_check_mtu flags*/ 6928enum bpf_check_mtu_flags { 6929 BPF_MTU_CHK_SEGS = (1U << 0), 6930}; 6931 6932enum bpf_check_mtu_ret { 6933 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 6934 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6935 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 6936}; 6937 6938enum bpf_task_fd_type { 6939 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 6940 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 6941 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 6942 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 6943 BPF_FD_TYPE_UPROBE, /* filename + offset */ 6944 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 6945}; 6946 6947enum { 6948 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 6949 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 6950 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 6951}; 6952 6953struct bpf_flow_keys { 6954 __u16 nhoff; 6955 __u16 thoff; 6956 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6957 __u8 is_frag; 6958 __u8 is_first_frag; 6959 __u8 is_encap; 6960 __u8 ip_proto; 6961 __be16 n_proto; 6962 __be16 sport; 6963 __be16 dport; 6964 union { 6965 struct { 6966 __be32 ipv4_src; 6967 __be32 ipv4_dst; 6968 }; 6969 struct { 6970 __u32 ipv6_src[4]; /* in6_addr; network order */ 6971 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6972 }; 6973 }; 6974 __u32 flags; 6975 __be32 flow_label; 6976}; 6977 6978struct bpf_func_info { 6979 __u32 insn_off; 6980 __u32 type_id; 6981}; 6982 6983#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6984#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6985 6986struct bpf_line_info { 6987 __u32 insn_off; 6988 __u32 file_name_off; 6989 __u32 line_off; 6990 __u32 line_col; 6991}; 6992 6993struct bpf_spin_lock { 6994 __u32 val; 6995}; 6996 6997struct bpf_timer { 6998 __u64 :64; 6999 __u64 :64; 7000} __attribute__((aligned(8))); 7001 7002struct bpf_dynptr { 7003 __u64 :64; 7004 __u64 :64; 7005} __attribute__((aligned(8))); 7006 7007struct bpf_list_head { 7008 __u64 :64; 7009 __u64 :64; 7010} __attribute__((aligned(8))); 7011 7012struct bpf_list_node { 7013 __u64 :64; 7014 __u64 :64; 7015} __attribute__((aligned(8))); 7016 7017struct bpf_rb_root { 7018 __u64 :64; 7019 __u64 :64; 7020} __attribute__((aligned(8))); 7021 7022struct bpf_rb_node { 7023 __u64 :64; 7024 __u64 :64; 7025 __u64 :64; 7026} __attribute__((aligned(8))); 7027 7028struct bpf_refcount { 7029 __u32 :32; 7030} __attribute__((aligned(4))); 7031 7032struct bpf_sysctl { 7033 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7034 * Allows 1,2,4-byte read, but no write. 7035 */ 7036 __u32 file_pos; /* Sysctl file position to read from, write to. 7037 * Allows 1,2,4-byte read an 4-byte write. 7038 */ 7039}; 7040 7041struct bpf_sockopt { 7042 __bpf_md_ptr(struct bpf_sock *, sk); 7043 __bpf_md_ptr(void *, optval); 7044 __bpf_md_ptr(void *, optval_end); 7045 7046 __s32 level; 7047 __s32 optname; 7048 __s32 optlen; 7049 __s32 retval; 7050}; 7051 7052struct bpf_pidns_info { 7053 __u32 pid; 7054 __u32 tgid; 7055}; 7056 7057/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7058struct bpf_sk_lookup { 7059 union { 7060 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7061 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7062 }; 7063 7064 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7065 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7066 __u32 remote_ip4; /* Network byte order */ 7067 __u32 remote_ip6[4]; /* Network byte order */ 7068 __be16 remote_port; /* Network byte order */ 7069 __u16 :16; /* Zero padding */ 7070 __u32 local_ip4; /* Network byte order */ 7071 __u32 local_ip6[4]; /* Network byte order */ 7072 __u32 local_port; /* Host byte order */ 7073 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7074}; 7075 7076/* 7077 * struct btf_ptr is used for typed pointer representation; the 7078 * type id is used to render the pointer data as the appropriate type 7079 * via the bpf_snprintf_btf() helper described above. A flags field - 7080 * potentially to specify additional details about the BTF pointer 7081 * (rather than its mode of display) - is included for future use. 7082 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7083 */ 7084struct btf_ptr { 7085 void *ptr; 7086 __u32 type_id; 7087 __u32 flags; /* BTF ptr flags; unused at present. */ 7088}; 7089 7090/* 7091 * Flags to control bpf_snprintf_btf() behaviour. 7092 * - BTF_F_COMPACT: no formatting around type information 7093 * - BTF_F_NONAME: no struct/union member names/types 7094 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7095 * equivalent to %px. 7096 * - BTF_F_ZERO: show zero-valued struct/union members; they 7097 * are not displayed by default 7098 */ 7099enum { 7100 BTF_F_COMPACT = (1ULL << 0), 7101 BTF_F_NONAME = (1ULL << 1), 7102 BTF_F_PTR_RAW = (1ULL << 2), 7103 BTF_F_ZERO = (1ULL << 3), 7104}; 7105 7106/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7107 * has to be adjusted by relocations. It is emitted by llvm and passed to 7108 * libbpf and later to the kernel. 7109 */ 7110enum bpf_core_relo_kind { 7111 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7112 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7113 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7114 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7115 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7116 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7117 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7118 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7119 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7120 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7121 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7122 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7123 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7124}; 7125 7126/* 7127 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7128 * and from libbpf to the kernel. 7129 * 7130 * CO-RE relocation captures the following data: 7131 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7132 * its insn->imm field to be relocated with actual field info; 7133 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7134 * type or field; 7135 * - access_str_off - offset into corresponding .BTF string section. String 7136 * interpretation depends on specific relocation kind: 7137 * - for field-based relocations, string encodes an accessed field using 7138 * a sequence of field and array indices, separated by colon (:). It's 7139 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7140 * arguments for identifying offset to a field. 7141 * - for type-based relocations, strings is expected to be just "0"; 7142 * - for enum value-based relocations, string contains an index of enum 7143 * value within its enum type; 7144 * - kind - one of enum bpf_core_relo_kind; 7145 * 7146 * Example: 7147 * struct sample { 7148 * int a; 7149 * struct { 7150 * int b[10]; 7151 * }; 7152 * }; 7153 * 7154 * struct sample *s = ...; 7155 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7156 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7157 * // b is field #0 inside anon struct, accessing elem #5) 7158 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7159 * 7160 * type_id for all relocs in this example will capture BTF type id of 7161 * `struct sample`. 7162 * 7163 * Such relocation is emitted when using __builtin_preserve_access_index() 7164 * Clang built-in, passing expression that captures field address, e.g.: 7165 * 7166 * bpf_probe_read(&dst, sizeof(dst), 7167 * __builtin_preserve_access_index(&src->a.b.c)); 7168 * 7169 * In this case Clang will emit field relocation recording necessary data to 7170 * be able to find offset of embedded `a.b.c` field within `src` struct. 7171 * 7172 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7173 */ 7174struct bpf_core_relo { 7175 __u32 insn_off; 7176 __u32 type_id; 7177 __u32 access_str_off; 7178 enum bpf_core_relo_kind kind; 7179}; 7180 7181/* 7182 * Flags to control bpf_timer_start() behaviour. 7183 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7184 * relative to current time. 7185 */ 7186enum { 7187 BPF_F_TIMER_ABS = (1ULL << 0), 7188}; 7189 7190/* BPF numbers iterator state */ 7191struct bpf_iter_num { 7192 /* opaque iterator state; having __u64 here allows to preserve correct 7193 * alignment requirements in vmlinux.h, generated from BTF 7194 */ 7195 __u64 __opaque[1]; 7196} __attribute__((aligned(8))); 7197 7198#endif /* _UAPI__LINUX_BPF_H__ */